Bioprocess Technology Kinetics and Reactors by Professor Dr. Anton Moser

Bioprocess Technology
t1ETHODOl06ICAl
ASPECTS
ethics
ecology
modeling
design &. development
economy
quantification
UP STREAM ENZYME ENGINEERING
BIOREACTOR PERFORMANCE FERMENTATION ENGINEERING
DOWN STREAM multj-S I multi X -processing
MEASURE &. CONTROL PLANT &. ANIMAL CELL
ENGINEERING
FIElD or
APPlICA TlON
BIOPROCESS TECHNOLOGV
Engineering Disciplines in Biotechnologies

Health care (Pharma)
Agro &. Food
Environmental
Industrial
Anton Moser
Bioprocess Technology
Kinetics and Reactors
Revised and Expanded Translation
Translated by Philip Manor
With 279 Illustrations
Springer-Verlag
New York Wien
Professor Dr. ANTON MOSER
Institut fUr Biotechnologie, Mikrobiologie und Abfalltechnologie,
Technische Universitat Graz, A-SOlO Graz, Austria
Translator
Dr. PHILIP MANOR
Library of Congress Cataloging-in-Publication Data

Moser, Anton, Dipl.-Ing. Dr. techno
Bioprocess technology.
Rev. and translated from the German ed.:
Bioprozesstechnik: Berechnungsgrundlagen der Reaktionstechnik biokatalytischer Prozesse.
Bibliography: p.
Includes index.
1. Biochemical engineering. I. Title.
TP248.3.M6713 1988 660'.63 87-26590
Printed on acid-free paper.
Revised and expanded translation from the German edition

Bioprozesstechnik: Berechnungsgrundlagen der Reaktionstechnik biokatalytischer Prozesse.
1981 by Springer-Verlag Wien.
ISBN -13 :978-1-4613-8750-3
1988 by Springer-Verlag New York Inc.
Softcover reprint of the hardcover 1st edition 1988
All rights reserved. This work may not be translated or copied in whole or in part without the
written permission of the publisher (Springer-Verlag, 175 Fifth Avenue, New York, NY 10010,
USA), except for brief excerpts in connection with reviews or scholarly analysis. Use in connection
with any form of information storage and retrieval, electronic adaptation, computer software,
or by similar or dissimilar methodology now known or hereafter developed is forbidden.
The use of general descriptive names, trade names, trademarks, etc. in this publication, even if
the former are not especially identified, is not to be taken as a sign that such names, as understood
by the Trade Marks and Merchandise Marks Act, may accordingly be used freely by anyone.
Typeset by Asco Trade Typesetting Ltd., Hong Kong.
9 8 7 6 543 2 1
ISBN -13:978-1-4613-8750-3 e- ISBN -13 :978-1-4613-8748-0

DOl: 10.1007/978-1-4613-8748-0
Preface
This book is based on a 1981 German language edition published by Springer-

Verlag, Vienna, under the title Bioprozesstechnik. Philip Manor has done the
translation, for which I am deeply grateful.
This book differs from the German edition in many ways besides language.
It is substantially enlargened and updated, and examples of computer simula-
tions have been added together with other appendices to make the work both
more comprehensive and more practical.
This book is the result of over 15 years of experience in teaching and
research. It stems from lectures that I began in 1970 at the Technical University
of Graz, Austria, and continued at the University of Western Ontario in
London, Canada, 1980; at the Free University of Brussels, 1981; at Chalmers
Technical University in G6teborg, Sweden; at the Academy of Sciences in
lena, East Germany; at the "Haus der Technik" in Essen, West Germany,
1982; at the Academy of Science in Sofia, Bulgaria; and at the Technical
University of Delft, Netherlands, 1986.
The main goals of this book are, first, to bridge the gap that always exists
between basic principles and applied engineering practice, second, to enhance
the integration between biological and physical phenomena, and, third, to
contribute to the internal development of the field of biotechnology by
describing the process-oriented field of bioprocess technology.
To achieve these goals, I have attempted to unify the many interdisciplinary
fields that continue to become ever more specialized. For better comprehen-
siveness, quantitative facts are often followed by a qualitative discussion but
without complete derivation of the final equations-these can be found in
the original references. Since a major goal is to contribute to the development
of a biotechnical methodology by integrating several fields of science into one,
details are often omitted in that they are thought to be not significant for
the overall concept of integrating strategy. This approach aims to foster new
orientation, a unified, process-oriented strategy of "bioprocess technology"
that follows a systematic method of thinking and working. This strategy,
as an expression of research philosophy, includes four working principles:
(a) working with simplifica.tions, that is-making distinctions between the
vi Preface
important and the unimportant, (b) quantifying, (c) process (regime) analysis,
that is, separating biological from physical phenomena, and (d) thinking and
working in terms of mathematical models.
In conjunction with the irreplaceable element of intuition, which always
serves as the starting point, mathematical models should be considered as
working hypotheses that can assist process development. Within the frame-
work of adaptive model building, these models can be compared with, and
fitted to, the experimental realit-y of biological processes.
The book is organized into chapters that seem sensible from a pedagogical
viewpoint. After the defitrltion of bioprocess technology and its delineation
with respect to the whole area of biotechnology in Chapter 1, and description
of principles of thinking and working to be used in the integration in Chapter
2, Chapters 3 through 5 discuss bioprocess analysis. Chapter 3 characterizes
bioreactors quantitatively, as does Chapter 4, which also covers the general
utilization of bioreactors for obtaining kinetic data in the analysis of specific
processes. In Chapter 5 the use of mathematical models for simulating the
kinetics of biological processes is described. Chapter 6 discusses the synthesis
of biological data (kinetics) and physical data (transport phenomena) in order
to estimate bioreactor performance, that is, the conversion and productivity
of the most important types and operations of reactors.
This book does not primarily provide an exhaustive survey of the literature,
but concentrates on the most significant facts. Rather, its principal feature is
its description of a generally valid approach to calculations for bioprocess
technology. The problems of achieving a quantitative understanding are
primary, especially the problems of understanding kinetics as a cycle of
engineering considerations and calculations. The biological synthesis of the
products of intermediary metabolism and all biochemicaljbiological aspects
are excluded; they are well described in other books, for example, volumes 1
through 8 of Biotechnology-A Comprehensive Treatise (Rehm, H.J., and
Reed, G., eds., 1982ff) and Comprehensive Biotechology (Moo-Young, 1985).
Discussion of the technical aspects of quantifying reactors and process develop-
ment is limited in favor of description of working procedures for the analysis
and synthesis of production-scale biological processes. As a consequence of
the novel strategy being developed here, I hope to close a gap I perceive in
the literature, especially for students and for people involved in industry.
The title "Bioprocess Technology" for this edition has been chosen in
accordance with the semantic definition of "technology" as the "science of
industrial arts" This title was preferred to the "working title" "Bioprocess
Engineering." (Another possibility was "Bioprocess Engineering Sciences.")
I believe that the conception of engineering basically means the area of applied
sciences, that is, execution of the science of industrial arts, including mechan-
ical aspects, auxiliary equipment, and measurement and control techniques,
perhaps even being dominated by them (cf. title of the journal Bioprocess
Engineering).
Preface Vll
This book attempts to present the formal kinetic macro-approach following

Einstein's dictum that "Everything should be made as simple as possible, but
not simpler." And I firmly believe that "Nothing is more practical than theory,
and-at the same time-nothing is more theoretical than practice." as a
typical example if the "Yin & Yang" principle, which is at present stressed by
Fritjof Capra in order to achieve a holistic, ecological paradigm.
Graz, Austria ANTON MOSER

Contents
Preface .................................................. . v
Nomenclature (List of Symbols) ............................. . xv
Subscripts ............................................. . xxv
Abbreviations .......................................... . XXVll
Greek Symbols ......................................... . xxviii
Chapter 1 Introduction ................................... .

1.1 Biotechnology: A Definition and Overview .................. . 1
1.2 Bioprocess Technology ................................. . 5
Bibliography ......................................... . 12
Chapter 2 The Principles of Bioprocess Technology ........... . 13

2.1 Empirical Pragmatic Process Development .................. . 13
2.1.1 Production Strains .................................... . 13
2.1.2 Starting Points ....................................... . 13
2.1.3 Different Modes (or Strategies) in Process Development ........ . 15
2.1.4 Process Development Without Mathematical Models .......... . 15
2.2 Basics of Quantification Methods for Bioprocesses ............ . 18
2.2.1 Concepts of a Uniform Nomenclature for Bioprocess Kinetics .... . 19
2.2.2 The Rates of a Bioprocess ............................... . 20
2.2.3 Stoichiometry and Thermodynamics ....................... . 25
2.2.4 Productivity, Conversion, and Economics (Profit) ............. . 38
2.3 Systematic, Empirical Process Development with Mathematical
Models ............................................ " 41
2.3.1 An Integrating Strategy-A Basis for Biotechnological
Methodology ........................................ . 41
2.3.2 Working Principles of Bioprocess Technology ................ . 46
2.4 Mathematical Modeling in Bioprocessing ................... . 49
2.4.1 General Remarks ..................................... . 49
2.4.2 Model Building ....................................... . 51
2.4.3 Different Levels and Types of Kinetic Models ................ . 54
Bibliography ......................................... . 62
x Contents
Chapter 3 Bioreactors....................................... 66
3.1 Overview: Industrial Reactors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
3.1.1 Microbiological Reactors (Fermenters, Cell Tissue Culture Vessels,
and Waste Water Treatment Plants) ......................... 66
3.1.2 Enzyme Reactors. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
3.1.3 Sterilizers. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
3.2 Systematics of Bioreactors ................................ . 69
3.2.1 Homogeneous Versus Heterogeneous Systems ................. . 70
3.2.2 Mixing Behavior ....................................... . 70
3.3 Quantification Methods .................................. . 73
3.3.1 Residence Time Distribution (RTD)-Macromixing ............ . 74
3.3.2 Micromixing .......................................... . 81
3.3.3 Oxygen Transfer Rate (OTR) .............................. . 90
3.3.4 Degree of O 2 Utilization, '102 99
3.3.5 Degree of Hinterland, HI ................................. . 99
3.3.6 Power Consumption, P .................................. . 100
3.3.7 O 2 Efficiency (Economy) E02 101
3.3.8 Heat Transfer Rate, Hv TR ................................ . 101
3.3.9 Characteristic Diameter of Biocatalytic Mass J;, ............... . 105
3.3.10 Comparison of Process Technology Data for Bioreactors ........ . 105
3.3.11 Biological Test Systems .................................. . 110
3.4 Operational Modes and Bioreactor Concepts ................. . 112
3.5 Bioreactor Models ...................................... . 118
3.5.1 Modell: The Ideal Discontinuous Stirred Tank Reactor
(DCSTR) ............................................. . 119
3.5.2 Model 2: The Ideal Continuous Stirred Tank Reactor (CSTR)
with V = Constant ..................................... . 119
3.5.3 Model 3: The Ideal Semicontinuous Stirred Tank Reactor (SCSTR)
with V = Variable ...................................... . 119
3.5.4 Model 4: The Ideal Continuous Plug Flow Reactor (CPFR) or
Tubular Reactor ....................................... . 121
3.5.5 Model 5: The Real Plug Flow Reactor CPFR with Dispersion ..... . 122
3.5.6 Model 6: The Discontinuous Recycle Reactor (DCRR) .......... . 123
3.5.7 Model 7: The Continuous Recycle Reactor (CRR) .............. . 123
3.5.8 Multiple Phase Bioreactor Models .......................... . 124
3.6 "Perfect Bioreactors" in Bench and Pilot Scale for Process Kinetic
Analysis .............................................. . 126
Bibliography .......................................... . 130
Chapter 4 Process Kinetic Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . 138

4.1 Kinetic Analysis in Different Types of Reactors . . . . . . . . . . . . . . . . . 138
4.2 Regime Analysis-General Concept and Guidelines ............. 141
4.3 Test of Pseudo homogeneity . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . 146
4.4 Parameter Estimation of Kinetic Models with Bioreactors. . . . . . . . . 151
4.4.1 Integral and Differential Reactors ........................... 151
4.4.2 Integral and Differential Reactor Data Evaluation Methods ....... 154
Contents xi
4.4.3 Results of Differential and Integral Analysis: Linearization

Diagrams ............................................. . 157
4.5 Modeling Heterogeneous Processes .......................... . 168
4.5.1 External Transport Limitations ............................. . 171
4.5.2 Internal Transport Limitations ............................. . 175
4.5.3 Combined Internal and External Transport Limitations .......... . 183
4.5.4 Transport Enhancement .................................. . 188
4.5.5 Concluding Remarks ..................................... . 192
Bibliography ........................................... . 193
Chapter 5 Bioprocess Kinetics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197

5.1 Temperature Dependence, k(T), Water Activity, aw, and Enthalpy/
Entrophy Compensation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198
5.2 Microkinetic Equations Derived from the Kinetics of Chemical and
Enzymatic Reactions ..................................... . 204
5.2.1 The Dynamic Flow Equilibrium Approach to Life Processes ....... . 204
5.2.2 Contribution of Enzyme Mechanism to Bioprocess Kinetic Models .. . 206
5.2.3 Contribution of Chemical Kinetic Laws to Bioprocess Kinetic
Modeling ............................................. . 214
5.3 Basic Unstructured Kinetic Models of Growth and Substrate
Utilization (Homogeneous Rate Equations) .................... . 216
5.3.1 It = It(s): Simple Model Functions ofInhibition-Free Substrate
Limitation (Saturation-Type Kinetics) ........................ . 217
5.3.2 It = It(x): Influence of Biomass Concentration on Specific Growth
Rate ................................................. . 224
5.3.3 It = It(t): Extensions of Monod-Type Kinetics to Stationary and
Lag Phase ............................................. . 225
5.3.4 Negative Biokinetic Rates-The Case of Microbial Death and
Endogenous Metabolism .................................. . 227
5.3.5 Kinetic Model Equations for Inhibition by Substrates and Products .. 231
5.3.6 Kinetic Model Equations for Repression ...................... . 237
5.3.7 It = It(pH) ............................................. . 237
5.3.8 Kinetic Pseudo homogeneous Modeling of Mycelial Filamentous
Growth Including Photosynthesis ........................... . 237
5.3.9 Kinetic Modeling of Biosorption ............................ . 238
5.4 Kinetic Models for Microbial Product Formation ............... . 240
5.4.1 Metabolites and End Products ............................. . 240
5.4.2 Heat Production in Fermentation Processes ................... . 247
5.5 Multisubstrate Kinetics ................................... . 250
5.5.1 Sequential Substrate-Utilization Kinetics ..................... . 251
5.5.2 Simultaneous Substrate-Utilization Kinetics ................... . 253
5.5.3 Generalizations in Multisubstrate Kinetics .................... . 254
5.6 Mixed Population Kinetics ................................ . 259
5.6.1 Classification of the Types of Microbial Interactions ............. . 260
5.6.2 Kinetic Analysis of Microbial Interactions ..................... . 261
5.7 Dynamic Models for Transient Operation Techniques (Nonstationary
Kinetics) .............................................. . 272
xu Contents
5.7.1 Definitions of Balanced Growth and Steady-State Growth ........ 272

5.7.2 Mathematical Modeling of Dynamic Process Kinetics. . . . . . . . . . .. 274
5.8 Kinetic Models of Heterogeneous Bioprocesses . . . . . . . . . . . . . . . .. 283
5.8.1 Biofilm Kinetics. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 283
5.8.2 Unstructured Models of Pellet Growth ............. . . . . . . . . .. 288
5.8.3 Linear Growth ......................................... 290
5.9 Pseudo kinetics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 290
5.10 Kinetics of Sterilization. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 292
5.10.1 Basic Kinetic Approaches in Sterilization Kinetics. . . . . . . . . . . . . .. 292
5.10.2 Multicomponent Systems in Food Technology ................. 294
Bibliography. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 295
Chapter 6 Bioreactor Performance: Process Design Methods ... . .. 307

6.1 The Ideal Single-Stage, Constant-Volume Continuous Stirred Tank
Reactor, CSTR (Pseudo homogeneous L-Phase Reactor Model) . . . .. 308
6.1.1 Performance of the CSTR with Simple Kinetics. . . . . . . . . . . . . . . .. 308
6.1.2 Performance of the CSTR with Complex Kinetics. . . . . . . . . . . . . .. 311
6.1.3 Stability Analysis and Transient Behavior of the CSTR . . . . . . . . . .. 318
6.2 Variable Volume CSTR Operation (Fed-Batch and Transient
Reactor Operation) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 325
6.3 Multistage Single and Multistream Continuous Reactor Operation.. 329
6.3.1 Classification. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 329
6.3.2 Potentialities of Multistage Systems. . . . . . . . . . . . . . . . . . . . . . . . .. 330
6.3.3 Single-Stream Multistage Operation. . . . . . . . . . . . . . . . . . . . . . . . . 331
6.3.4 Multistream Multistage Operation .......................... 334
6.4 Continuous Plug Flow Reactors (CPFR) . . . . . . . . . . . . . . . . . . . . .. 337
6.4.1 Performance Equations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 337
6.4.2 Potential Advantages of CPFR Operation. . . . . . . . . . . . . . . . . . . .. 339
6.4.3 Principal Properties and Design of CPFRs Compared with CSTRs .. 340
6.4.4 Applications ofCPFR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 347
6.4.5 One-Phase (Liquid) Reactors with Arbitrary Residence Time
Distribution and Micromixing . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 347
6.5 Recycle Reactor Operation ................................ 351
6.5.1 Performance Equations of Recycle Reactors. . . . . . . . . . . . . . .. . .. 351
6.5.2 Application of CRR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 356
6.6 Gas/Liquid (Two-Phase) Reactor Models in Bioprocessing ........ 357
6.7 Biofilm Reactor Operation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 358
6.7.1 Potentialities of Biofilm Reactors. . . . . . . . . . . . . . . . . . . . . . . . . . .. 359
6.7.2 Performance Equations of Biofilm Reactors. . . . . . . . . . . . . . . . . . .. 360
6.7.3 Application of Biofilm Reactors. . . . . . . . . . . . . . . . . . . . . . . . . . . .. 370
6.8 Dialysis and Synchronous Culture Operation .................. 371
6.8.1 Dialysis (Membrane) Reactor Operation. . . . . . . . . . . . . . . . . . . . .. 371
6.8.2 Synchronous Culture Operation ............................ 378
6.9 Integrating Strategy as General Scale-Up Concept in Bioprocessing. 381
6.9.1 Stoichiometry (Balancing Methods) Applied in Bioprocess Design. .. 382
6.9.2 Interactions Between Biology and Physics via Viscosity of
Fermentation Media . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 385
Contents xiii
6.9.3 Influence of Mycelium-The Morphology Factors {"Apparent

Morphology") . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 389
6.9.4 Structured Modeling of Bioreactors (OTR) . . . . . . . . . . . . . . . . . . . .. 392
6.10 Final Note ............................................. 395
Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 396
Appendix I Fundamentals of Stoichiometry of Complex

Reaction Systems .............................. . 406
Appendix II Computer Simulations .......................... . 412
Appendix III Microkinetics: Derivation of Kinetic Rate
Equations from Mechanisms. . . . . . . . . . . . . . . . . . . . .. 436
Index..... .... .............. .... .. . . . ...... ...... ........ .. 441

Nomenclature (List of Symbols)
ARABIC
A [m 2 ] surface area
[A] Ampere
A,B,C,D compounds (general)
{A} activated state of
compound A
a [m-I] specific area
a, b, c, d, e, f coefficients (esp. in
element-species matrix,
cf. Equ. 11.6)
a, d, c, d [kgm- 3 ] concentration of
compounds
aLl [m 2 m- 3 ]
aw water activity (Equ. 5.1)
BOD [kgm- 3 C I ] biological oxygen

demand
Bv [kgm- 3 h- l ] volumetric mass loading
rate (Equ. 6.129)
Bx [h- I ] mass loading rate per
unit sludge biomass
(Equ. 5.212)
Bi Biot number
(Bi = D L2 bLZ/D.. bs )
Bo Bodenstein number
(Bo = vL/Dd
C [kg] amount of CO2

C [$] costs
C integration constant
xvi Nomenclature (List of Symbols)
CTD circulation time

distribution
c, Ci , cj [kg m- 3 ] concentration (general)
of component i or j
C [kg m- 3 ] concentration of CO 2
CIL constant in Equ. 5.218
Cp [kJ' kg- 1 . C- I ] specific heat at
p = constant
CR [kg m- 3 ] concentration of
ribosomes
Cst [n j 'm- 3 ] sterilization level
D [m 2 's- l ] diffusion or dispersion

coefficient
D [h- 1 ] dilution rate
Dc [h- I ] critical dilution rate
where washout occurs
DM [h- I ] dilution rate with
maximum productivity
DIO [h] decimal reduction time
(Equ.5.8)
Da (Da o , DaLls, Daold Damkoehler number
(Equs. 4.3,4.5,4.7)
Da( Damkoehler number of
first degree (Equ. 6.89)
Dan Damkoehler number of
second degree (Equ. 4.74)
d em] diameter (general term)
d constant
d em] mean diameter
dB em] bubble diameter
dj em] impeller diameter
dp em] particle diameter
dp em] active particle diameter
(Equ. 5.258)
dT em] tank diameter
E [m 2 's- 1 ] convection coefficient

(Equ. 3.30d)
E (E OTR ) enhancement factor
(Equ. 4.115) of OTR
E a , Eon Ed [J. mole-I] activation energy,
general (growth or
death)
Nomenclature (List of Symbols) xvii
[kg O 2 . kWh- 1 ] oxygen economy

(Equ. 3.66) (oxygen
efficiency)
E, ES, EP, EI [kg] amount of enzyme,
enzyme-substrate,
enzyme-product,
enzyme-inhibitor
complex
{ES} activated state of
ES-complex
e, es, ep, ei concentration of
components
[m 3 h- 1 ] volumetric flow rate

[m 3 h- 1 ] bleed rate
[m 3 h- 1 ] feed rate
[m 3 h- 1 ] pumping capacity
[kgm- 3 . h- 1] substrate feed rate
(Equ.11.30)
Fr Froude number
(Fr = v 2 /g. L)
F(t) mathematical RTD
function acc. to step
method
f(t) mathematical RTD
function acc. to pulse
method
singular f(t) (Equ. 3.14)
kinetic term for
overlapping multi-
substrate degradation
f2 kinetic term for
repression of substrate
degradation by other
substrates
fo' correction factor for
oxygen saturation
calculation (Equ. 3.40)
G Material in
"G compartment"
proteins and
DNA)
Ga Galilei number
XVlll Nomenclature (List of Symbols)
AG [J. mole- 1 ] free reaction enthalpy

g [cm S-2] acceleration of gravity
(980)
universal gravitational
constant (6.671 . 10- 8 )
H Em] height
H [m 3 atm mol- 1 ] Henry distribution
coefficient
He dimensionless Henry
distribution coefficient
Ha Hatta number
(Equ.4.112)
HI degree of hinterland
(Equ.3.59)
reaction enthalpy
(molar)
[J. kg-l] reaction enthalpy per
unit mass of X or S
heat of combustion of
components
amount of volumetric
heat of fermentation
h [J. s] Planck constant
(6.6. 10- 34 )
[Jm- 2 h- 1 K- 1 ] heat transfer coefficient
[kgm- 3 ] concentration of H+
ions
crowding factor
(Equ.6.188)
"concentration" of
volumetric heat
I amount of inhibitor
concentration of
inhibitor
increments in Equ. 3.34
J degree of segregation
(Equ.3.1)
K material in "K
compartment" (RNA
mainly)
Nomenclature (List of Symbols) xix
K [kgm- 3 ] constant of equilibrium

or saturation (general
term)
Kadapt constant of metabolic
adaptation in Equ. 5.227
Kc [(Nsm- Z )1/2] constant in Casson law
(Equ. 6.166), "Casson-
viscosity"
Ko constant diffusional
resistance in Equ. 5.43
Ke constant for endogenous
metabolism in Equ. 5.87
Keq constant of thermo-
dynamic equilibrium
(Equ. 5.22) "Haldane
relationship"
Kd constant for microbial
death (Equ. 5.86)
KIS' KIP (KISX' KIPX) [kgm- 3 ] inhibition constant of
growth by substrate or
product
K m , K s, K L , K F , KH [kgm- 3 ] saturatien constant for
substrate in kinetic
equation acc. to
Michaelis-Menten,
Monod, Langmuir,
Freundlich, and Hill
Ko,Kp [kgm- 3 ] saturation constant in
Monod-type kinetics for
oxygen or product
Ko constant in Equ. 3.34
KR (KISP) [kgm- 3 ] constant for repression
K 1 ,K z constants for pH
function (Equ. 5.107)
k [(m 3 mol- 1 )n-1 . S-l] rate constant (general
term)
k number of elements for
elemental balancing
method (Sect. 2.4.2.2)
kB [J. K- 1 ] Boltzmann constant
(1.38' IO- Z3 )
k cat [mole P' mole E- 1 S-l] catalytic constant or
turnover number
kd [h- 1 ] specific dead rate
xx Nomenclature (List of Symbols)
kE [S-l J rate constant of

electrode response
ke [S-lJ rate constant of
environmental change
kH (h) [Jm- 2 h- 1 K- 1J true rate coefficient of
volumetric heat transfer
kHTR [Jh- 1 K- 1J rate coefficient of
volumetric heat transfer
(phenomenological)
km [S-l J first-order rate constant
for mixing (Equ. 3.23)
kLla (kLa) [h- 1J volumetric OTR coeffi-
cient (GIL interface)
kL2 [ms- 1J transfer coefficients at
LIS interface
kp [h- 1J rate constant of product
formation
kp,d [h- 1J specific rate constant of
product degradation
kp1 , kp2 [h- 1J rate constant acc. to
Kono equation
(Equ. 5.126)
kr (k j , k n) [e 1 (mole/vol)-n+1 J reaction rate constant
(ith reaction or for
reaction order n)
ks [m' S-l J transport rate coefficient
in solid phase
kTR [S-l J transport rate constant
(phenomenological)
kv,G [h- 1J transport rate constant
of gas phase (Equ. 3.38)
kx [h-1J constant for biomass in
Equ. 5.63 or 6.170
k o, k 1, k1/2 [!J rate constants for reactor
order n = 0, 1, 1/2
k 1, k 2, k3 [!J rate constants in BRE
(Equ. 5.239), esp.
Equ. 4.25 with 5.240
k+ 1, k-1 [!J rate constants for
forward or backward
reaction
kt [h- 1J decay coefficient
kco [m 3. mole-I. S-l J preexponential factor in
Arrhenius equation
(Equ.5.3)
Nomenclature (List of Symbols) XXI
L em] length (general term)

L e, L t, L u , L H , em] morphology factors
(e.g., Equ. 6.182, 5.109)
M [kg] mass
MF morphology factor
(Equ. 6.183)
[h- 1 ] specific maintenance
Ins, Ina, InH,

rate coefficient for S, 2 ,
and Hv
In [%] degree of mixing
(Equ.3.17)
In power factor in
Equ. 6.164, "flow index"
(see also Equ. 6.171)
N number of significant
components in balancing
method (Sect. 2.4.2.2);
number of vessels in
senes
NA aeration number
(Equ.3.75)
Ni number of molecules of
component i
NM mixing number
(Equ. 3.73)
NOTR OTR number (Equ. 3.76)
Np power number
(Equ. 3.74)
NRe (Re) Reynolds number
(Re = V" Llv)
Nsc (Sc) Schmidt number
(Sc = vlD)
NSh (Sh) Sherwood number
(Sh L = k L LID)
n [rps] rotational speed, stirrer
speed
n (n., no) reaction order (with
respect to S or 02)
n power index
n number of microbial
cells
ni number (of moles of
component i)
xxii Nomenclature (List of Symbols)
nB bed expansion index

(Equ.6.139)
nH Hill coefficient
(Equ.5.29)
nj [kgm- 3 s- 1 ] volumetric flux of mass
(transport rate)
n~I [kgm- 2 . S-1] (mass) flux through area
0 kg amount of O2
OTR [kgm- 3 h- 1 ] O2 transfer rate
o (resp. 0*) [kgm- 3 ] O2 concentration (resp.
saturation value)
obs observed value
OUR [kgm- 3 h- 1 ] 02-uptake rate
P [kg] amount of product

P [WorJs- 1 ] power
Pmb [m 3 . h- 1 kN- 1 ] membrane permeability
(Equ.6.146a)
p [kgm- 3 ] concentration of
product
p [bar] pressure
Q product formation
activity function
(Equ. 5.220)
q [J m- 3 . S-1] rate of heat transfer
qj (q., qo' qc' qHv' qp) [h- 1 ] resp. specific rate of formation
J. kg- 1 h- 1 (for qH) or consumption of
component i (substrate,
O2, CO2, heat, and
product)
rate of internal
production in biomass
compartment
[J mol- 1 K- 1 ] gas constant (8.314.10 7 )
Em] radius
rank of matrix (general,
of stoichiometrical
coefficients and element-
species), see App. I
Reynold number (NRe )
terminal settling
velocity, Re-number of
bioparticles (Equ. 6.139)
Nomenclature (List of Symbols) xxiii
Re-number based on
energy dissipation
(Equ.4.1b)
RTD residence time
distribution
r excretion rate
r correlation coefficient
r recycle ratio
rate of production or
consumption
(volumetric)
rate of CO 2-production
rate of production
rate of S-consumption
rate of growth
rate of 02-consumption
[lm- 3 h- 1 ] rate of heat production
[kgm- 2 s- 1 ] rate per unit area
[S-l ] related reaction rate
(general term)
S [kg] amount of substrate

Sc Schmidt number (Nsc )
Sh Sherwood number (NSh )
IlS [J. mole- 1 . K- 1 ] reaction entropy
s crowding factor
(Equ. 6.183b) for space
filling
s concentration of
substrate
s surface renewal rate
(Equ.3.30)
standard deviation
"sludge loading" =
adsorbed S concentra-
tion (biosorption)
(Equ. 5.114)
critical threshold
concentration (Equ. 5.95)
concentration of
essential and enhancing
substrates (Equ. 5.168)
"washout state" S
concentration (cf.
Fig. 6.1a, 6.4, and 6.5)
xxiv Nomenclature (List of Symbols)
2S [kgm- 3 J "non washout state" S

concentration (steady
state)
3S
[kgm- 3 J unstable S concentration
(Fig. 6.11)
T [Oe,KJ temperature
Tp [KJ isokinetic temperature
(Equ.5.11)
TOe total organic carbon
[sJ time (general term)
t [hJ mean residence time
tc [sJ circulation time
tG [sJ mean residence time of
gas phase
t HTR [sJ characteristic time of
heat transfer (Table 4.3a)
tL [sJ lag time
tM [sJ maturation time
(Equ.5.136)
tm (tm,9S) [sJ mixing time
(at m = 95%)
t max [sJ time of maximum
production rate
(Equ.5.124)
tOTR [sJ characteristic time of
OTR (Table 4.1)
tp [sJ period of oscillation
(Equ.5.208)
tq; [sJ characteristic time of
consumption or
formation rates (cf.
Table 4.1)
tr [sJ characteristic reaction
time (cf. Table 4.1)
tSt [sJ sterilization time
V [m 3 J volume
[VJ Volt
v [m' S-l J velocity (general term)
v (see r) [kg'm- 3 's- 1 J reaction rate, esp. of
enzyme kinetics
VSL> VSG [m's- 1 J superficial velocity of
liquid or gas phase
Vt [m' s- l J terminal settling velocity
Nomenclature (List of Symbols) xxv
peripheric velocity of
stirrer (stirrer top
velocity)
v rnf velocity of minimum
fluidization
Vms velocity of minimum
spouting
w [kg] weight
[W] Watt
w mass fraction of
component
x amount of biomass
x biomass concentration
measurement points or
significant process
variables or arbitrary
quantity
"washout" and "non-
washout state" of
biomass concentration
(Fig.6.1a)
unstable steady state
(Fig. 6.11) biomass
concentration
y, 1';u yield coefficient (general

term) (see Table 2.14)
Z value (Equ. 5.9)

electric charge
(Equ.3.33)
z coordinate
SUBSCRIPTS
ads, a adsorption B bulk or bubble or

agit agitation bleed
app apparent BC bubble column
ass assimilation bio biological
asym asymptotic
adap adaptation C CO 2
av available c circulation
ave average c cold
xxvi Nomenclature (List of Symbols)
calc calculated L liquid phase

cat catalytic L longitudinal
charact characteristic L1 liquid film at GIL
crit, c critical interface
L2 liquid film at LI S
d death or decay interface
dc (DC) discontinuous
degr, M maturation
degrad, d degradation m mixing degree,
dist distributor medium, membrane
mm maximum mixedness
E electrode or enzyme max maximum
or single mb membrane
e endogenous or mIll minimum
environmental or mf minimum fluidization
electrons ms minimum spouting
eff effective
elim, el, e elimination N number of stages
end end N resp. Ni nucleotide resp.
est estimated internal nucleotide
eq equivalent
evap evaporation (heat loss) 0 oxygen
ex exit 0 initial value
exp experimental obs observed
ext external opt optimal
OTR oxygen transfer rate
F feed
f film, biofilm II product sum
ferm fermenter p product
p power
G gas phase, gas pr production
G G compartment
gl glucose R reactor or ribosomes
gr growth r reaction or resistant
or recycle
H heat, general rad radiation
Hv volumetric heat rds rate-determining step
reI relative
1, J number, components repr repression
III inlet res reservoir
ingest ingestion
int internal L sum
intersect intersection S solid phase or sub-
strate
K K compartment ST stirred tank
Nomenclature (List of Symbols) xxvii
s sensitive COD chemical oxygen

salt salt demand
Slm simulated CPFR continuous plug flow
st stirrer or sterilisation reactor
CRR continuous recycle
T tank reactor
TR transfer CSTR continuous stirred tank
terminal reactor
tot total, global CTCR cycle tube cyclone
ts total segregation reactor
CTD circulation time
V volumetric distribution
v viable
DCRR discontinuous recycle
W wall reactor
w warm DCSTR discontinuous stirred
tank reactor
X biomass DNA desoxyribonucleic acid
z coordinate ES enzyme substrate

=I- value of activated complex
complex EFB European Federation
/\ normalized value of Biotechnology
mean value (steady E, dE 260 extinction esp. at
state) 260/lm
FBBR fluidized bed biofilm

ABBREVIATIONS
reactor
ADH, adh alcohol dehydrogenase FDP fructose diphosphate
ADP adenosine diphosphate
AMP adenosine mono- G G-compartment in
phosphate structured modeling
ATP adenosine triphosphate of cells
ATPase adenosine tripho- gapdh glyceraldehyde-3-
sphatease (enzyme) phosphate dehydro-
genase
BFF biological film Hv TR volumetric heat transfer
fermenter rate
BOD biological oxygen
demand HK hexokinase
BRE biological rate equation
idCPFR ideal continuous plug
CMMFF completely mixed flow reactor
microbial film idCSTR ideal continuous stirred
fermenter tank reactor
XXV111 Nomenclature (List of Symbols)
K K-compartment in rCSTR continuous stirred tank

structured modeling reactor with cell
of cells recycling
RF rotor fermenter
mm maximum mixedness rH redox potential
RNA ribonucleic acid
NAD nicotinamide adenine RTD residence time
dinucleotide distribution
NADH nicotinamide adenine
dinucleotide, reduced S Solid
form s substrate or solid
NCSTR cascade of CSTR with SE substrate enzyme
N stages complex
Su Sulfur
OTR O 2 transfer rate SCR semicontinuous reactor
OUR O 2 uptake rate SCSTR semicontinuous stirred
~O log mean of O 2 tank reactor
concentration SOY sum of variances
SLTR liquid-phase substrate
PATR pneumatically agitated transfer rate
& aerated tubular SsTR solid-phase substrate
reactor transfer rate
pfk pyruvate fructokinase
PGK phosphoglycerate TLR tubular loop reactor
kinase ThLTBFF thin-layer tubular
Ph Phosphorus biofilm fermenter
PK pyruvate kinase ThLTR thin-layer tubular
PHB polyhydroxy butyric reactor
acid ThLTFF thin-layer tubular film
Pr productivity fermenter
pyk pyruvate kinase TOC total organic carbon
ts total segregation
qss quasi-steady-state ~T log-mean of tempera-
q heat (rate) ture
TOC total organic carbon
rds rate-determining step TR transport, general
GREEK SYMBOLS
rx(varrx) vanance
rx, [3, y, b, coefficients
[3 concentration factor of settling device
[3 (o) modulus (Equ. 4.85)
y [!] factor in Equ. 4.123
y [S-1 ] shear rate
Nomenclature (List of Symbols) xxix
Ys,Yx degree of reductance (reduction)

() [mJ thickness of film acc. to two-film theory
()* morphology factor (Equ. 6.178)
e [m 2 s- 3 J energy dissipation
e ratio of Ks values (Equ. 5.137)
eG gas hold-up
ep volume fraction of particles
ex biomass hold-up (Equ. 6.134)
(i (relative) conversion of component i (Equ. 2.45)
'1 effectiveness factor (general term)
'1 [Nsm- 2 J dynamic viscosity
fj overall effectiveness factor (Equ. 4.2)
'1app [Nsm- 2 J apparent viscosity
'1D factor in Equ. 3.62
'1E energy efficiency coefficient of growth
(Sect. 2.2.3.4)
'1GIL effectiveness factor of GIL-mass transfer
'1Lls effectiveness factor of LI S-mass transfer
'1M factor in Equ. 3.60
'10 [Nsm- 2 J viscosity of suspending fluid (for l/Js --+ l/J)
'102 [%J O 2 utilization (Equ. 3.58)
'1r.A surface-area-related effectiveness factor of
reaction (Equ. 4.98)
'1r ('1r. v) effectiveness factor of reaction (volume related)
'1~ effectiveness factor of reaction (Equ. 4.108)
~r effectiveness factor of reaction (Eq u. 4.11 Ob)
'1th thermodynamic efficiency
'1TR(E) effectiveness factor of transport (enhancement
factor)
e [!J factor in Equ. 3.63
ex [hJ sludge age (Equ. 6.122)
/( dimensionless factor for bulk reaction rate in
Fig. 5.75
A [hJ cell age (Equ. 5.134)
A [n- 1 cm- 1 J conductivity
/1 [h- 1 J specific growth rate
fJ. [h- 2 J first derivation of specific growth rate on time
/1d [h- 1 J specific microbial death rate (Equ. 5.86)
/11 first moment (average value) of distribution
function (Equ. 6.191)
v [m 2 's- 1 J kinematic viscosity
v [S-l J specific doubling rate of cell number
Vi stoichiometric coefficients (Equ. 1.4)
~ [moleJ extent of reaction (Equ. 1.3)
~ads sorption capacity (Equ. 5.115)
xxx Nomenclature (List of Symbols)
II product sum
p density
~ sum
ionic strength (Equs. 3.33, 5.93)
selectivity (Equ. 2.43)
deviation (Equ. 4.47)
spread of distribution function (Equ. 6.191)
variance of distribution function
shear stress
normalized time (t/f)
[s] characteristic time of electrode response
[s] characteristic time of environmental change
[s] response to signal
[s] characteristic time of response acc. to moment
method with first-order lags in series (Equ. 351)
lag time (transients), Equ. 5.213
yield stress
Thiele modulus (Equs. 4.80 and 4.77) (internal
limitation)
modulus Equ. 5.251
modulus for zero-order reaction (Equ. 4.85)
modulus for first-order reaction (Equ. 4.83)
generalized Thiele modulus of particles
Thiele modulus based on biofilm thickness ()
(Equ.4.111)
Thiele modulus for external S limitation
(Equ.4.63)
Jlimit limiting value of lex! (Equ. 4.65)
consumption coefficient (Equ. 5.48)
Carman factor in Equ. 4.1a
volume fraction of suspended solids (Equ. 6.184)
CHAPTER 1
Introduction
1.1 Biotechnology: A Definition and Overview

In a general sense, biotechnology is a multi- and interdisciplinary field of
activities dealing with biological and biochemical processes carried out on
a broad range of scale (10 I to 106 m 3 ) as a technique for production.
Biotechnology has been defined differently in different countries and con-
tinents. Based on a definition by the European Federation of Biotechnology
(EFB), a broader sense is discussed in Europe recently:
"Biotechnology is the integrated use of natural sciences (biology inc!. molecular biology,
biochemistry, but also chemistry and physics) and engineering sciences (chemical
reaction engineering, electronics) in order to achieve the application of (technological,
industrial) organisms, cells (microbes, plant and animal cells), parts thereof and molec-
ular analogues in order to provide biosociety desirable goods and services"
Thus, the main characteristics of biotechnology are integration, orientation
towards applications but also a considerably impact on human society in the
future ("Biosociety", FAST, 1980; Moser, 1988).
Significant fields of applications are
Industrial Biotechnology
Food- and Agricultural (Phyto-, Zoo-) Biotechnology
Healthcare (Pharma-) Biotechnology and
Environmental Biotechnology.
This definition excludes medical engineering, which is concerned with the
construction of instruments for biological or health care use~for example,
heart-lung machines. Biomedical engineering is also commonly referred to as
biotechnology and/or bioengineering, and no clear-cut distinction is apparent
from the name alone. Further, technological activities in the field of agricul-
ture, for example, traditional breeding (phytotechnology and zootechnology),
belong to biotechnology only in the broadest sense.
One systematic treatment of biotechnology results from a logical considera-
tion of the catalysts responsible for the conversions in biochemical reactions
2 1. Introduction
and other biologically active agents. In this sense, biocatalysts are not only
naturally or synthetically produced enzymes and biomolecules respectively
analogues but also cells or subcellular fractions from microorganisms and
from plant and animal tissues. The type of catalyst used to carry out a bio-
logical process delineates the individual technology: enzyme or (simple and
complex) fermentation technology or tissue culture with suspended or im-
mobilized catalysts. In addition to the production techniques, there is also a
group of sterilization technique used to inactivate or kill biological materials.
For historical reasons, precise separation of biotechnology from other fields
is often difficult. There is overlap among various areas, such as conventional
food technology and biotechnology (Fig. 1.1). The production of yogurt or
TECHNICAL
B lOP ROC E SSE S
--------
(BIOTECHNOLOGY and FOOD TECHNOLOGY)
~
Production
of
.
Destructlon
of
biological material biological material
('JtS~zy~-
~~'I '\ "'" TECHNOLOGY
J
STERILIZATION-
TECHNOLOGY
FOOD BEVERAGE WASTE WATER-,WASTE- TISSUES-

PROCESSING INDUSTRY TECHNOLOGY CULTURE
e.g.yoghurt e.g. animal,plant
beer,wine cell s
t
t
, I
I
t
I I
~
I
,I t + + ~
MICROBIAL REACTORS ENZYME
t REACTORS STERILIZERS
VAT
BARREL
DRUi~
"jEHTER WAST~~:':,iR-~W_~~_l_~T_ :[~~~E- I
f
B lOR E C TOR S
L-----------~t+t~-------------~
H I 0 LOG I CAL REA C TOR S
FIGURE 1.1. Biological reactors and their areas of application for "taming" many
industrial-scale bioprocesses. (After Moser, 1978, 1981.)
1.1 Biotechnology: A Definition and Overview 3
cheese, for example, would be considered as bioprocesses of food technology.

The high degree of complexity of bioprocessing in food technology and
the difficulty of obtaining quantitative measurements (a difficulty shared by
many other divisions of fermentation technology) usually retard or inhibit
a systematic approach. In addition, special expectations exist for the quality
of the foodstuff resulting from bioprocessing and the techniques of large-scale
production are only slowly being introduced. Despite the fundamentally
different orientation of food technology, however, progress is being made in
introducing scientific methods. Because of commercial considerations, sooner
or later the working principles of biotechnology will be applied in complex
processes.
Precisely the same fate can be seen in neighboring area, the biological
treatment of waste water and solid wastes (Fig. 1.1). The pressure of environ-
mental problems has grown so great that here too the methods of thinking
that have been successfully used in conventional biotechnical processes are
being applied. One can now speak of the biotechnology of waste water
treatment.
Figure 1.1 shows the correspondence between the various technologies of
bioprocessing and the containers in which the reactions are carried out on
an engineering scale. These "bioreactors" range from sterilizers to proper
bioreactors (microbiological reactors such as fermenters, waste water treat-
ment units, vessels for cultivation of plant & animal cells, enzyme reactors) to
miscellaneous containers such as bottles, barrels, and so on used primarily for
carrying out "arts and crafts"-type complex biological processes.
What are the future prospects of biotechnology? The range of products
is being extended from antibiotics, steroids, vitamins, diagnostics, vaccines,
enzymes, and polysaccharides to such fine chemicals as organic acids, alcohols,
and single-cell protein (Rehm and Reed, 1982ff). Figure 1.2 shows the course
of development of prices in relation to the scale of production facilities for
the period 1970 to 2000 (Hines, 1980). From this figure one can conclude that,
at present, biotechnical methods are more economical than chemical ones
only in the case of pharmaceuticals that are complex molecules. The prediction
is that soon single-cell proteins can be economically produced by biotechnical
methods, and that by 1995 even inexpensive chemicals may be produced
in the same way. Economic feasibility in processes for producing simple
materials is first reached only in large-scale production units. At present,
small-scale, high-price processes are being carried out, mainly on a dis-
continuous basis. For the future, one looks toward large-scale, low-price
processes carried out on a continuous basis. Additional facts concerning
economics are discussed by Hamer (1985).
Some general conclusions can now be drawn from experience in fermentation,
food, and waste water treatment engineering:
1. The development of processes from the laboratory stage to the stage of tech-
nical maturity where they can be of service to mankind must be accelerated
and made more reliable. Three stages are involved in "taming" a bioprocess.
4 1. Introduction
PInJI{SS
IN
SElllrfJ
PRICE
==~~~-------------~t t
BlnTEDffiIlXiY
t PHARMACEUT ICALS
1970
........,...
-... - ...... --.:..-

:"'-:;:.:-;, ;.--"".-=r
1980
SIMPLE CHEMICALS ,~'Wi--_~- t 1990
b --,---
- PROIXJCTION SCALE
FIGURE 1.2_ Predicted development of biotechnical processes (---) compared with

development of chemical processes (--) for three classes of products (pharmaceuticals,
single-cell protein, simple chemicals) as examples of products with different molecular
complexity. The date (t) shows when it is predicted that the bioprocess will be more
economical than the chemical process. Region a: small-scale, high-price processes.
Region b: large-scale, low-price processes. (After Hines, 1980. With permission of
Butterworth scientific Ltd.)
The first is that of nature itself. In the earth, -in organisms, and in water,
biological processes operate on their own. They are often first discovered when
human intervention disturbs them. This stage was, and is, the school in which
human beings learn through trial and error to reproduce the processes and
use them for their own ends. The protocols for doing this are strictly observed
and are often kept secret. This was, and is, the stage of the artist and crafts-
person in, for example, the making of beer, wine, and sauerkraut. In part, it
is still the foundation upon which modern technology builds to create complex
products. Modern technology goes beyond this foundation to ensure repro-
ducibility of products through quantitative methods. In the field of chemical
engineering, such reproducibility was accomplished several decades ago.
2. Developing completely new methods of production demands a new organi-
zation of scientific and technical work. In recent years the problems of raw
material supplies, energy, and the environment have posed new questions.
With processes based on biotechnology, previously unused or underused
natural sources of energy and raw materials can be used or used more
intensively, and disturbed ecological cycles can be stabilized (for example,
CO2 , sunlight, cellulose, waste materials, and the cycles of the self-cleaning
rivers).
Industrial use of biotechnology requires new technical knowledge, new
1.2 Bioprocess Technology 5
working methods, and valid principles of operation. The new processes extend
well beyond the framework of traditional technology in both complexity and
scale. Neither simple fermentation technology nor waste water treatment
technology, nor the technologies of food processing, pharmaceuticals, or water
purification, provide a model adequate to serve as the basis for organization
of the new types of production. Traditional technologies developed largely in
isolation; they were product oriented. The current existence within each field
of independent terms describing technical measures and quantities common
to them all is a sign of this product orientation.
To foster the rapid and reliable development of processes based on bio-
technology, and especially to foster new production methods applicable on a
usable scale, a more unified view is needed. The many similar biotechnical
methods used by all of the scientific disciplines represented and by all of the
industries involved must be placed on a more universal basis, both from neces-
sity and from consideration of the advantages of universality. The product-
oriented way of thinking should be woven into a "process-oriented" viewpoint
that unites processes having similar working principles (Dechema, 1974; A.
Moser, 1977; Ringpfeil, 1977). In the chemical industry, such process orienta-
tion led to the development of chemical engineering. Figure 1.1 shows the
process-oriented way of thinking with regard to the food, beverage, and waste
water treatment industries. The different bioreactors must be reduced to a few
basic types. Finally, general working principles are needed. It is necessary and
advantageous to universalize concepts, definitions, and nomenclature.
1.2 Bioprocess Technology

The circumstances just described are the starting point of this book and the
initiative for the book's objectives. The term "bioprocess technology," created
by the author in 1972 includes the reaction engineering of biochemical pro-
cesses (Reuss, 1977). The term is now increasingly accepted for the integrated
disciplines of engineering sciences in biotechnology (i.e., kinetics, reactors,
measurement and control, upstream and downstream processing). Thus, it
includes the general working principles for introducing, executing, and opti-
mizing biological processes on a technical scale. The term "biotechnology"
fulfills its comprehensive role only when it is understood to mean the scientific
discipline of general applicability associated with engineering-scale biopro-
cesses made reproducible through the utilization of systematic, quantitative,
and system-analytical methods. The inter and trans-disciplinary character of
biotechnology that is reflected in the hybrid name is also found in other areas
of engineering. No doubt, the purely technical component of reactors and their
layout is fundamentally the same as in chemical engineering. The special
feature of biotechnology lies in the nature of the biological processes, in the
interaction of biochemistry and biology with technical capabilities. This book
reflects this special feature in that it discusses less the problems of reactors
6 1. Introduction
and more the problems involved in integrated quantification of bioprocesses.

The emphasis here, then, will be on bioprocess kinetics, because kinetics
represents, so to speak, the heart of bioprocess technology including inter-
actions with physics.
A further limitation on the contents of this book is the interdisciplinary
structure of biotechnology. Complete coverage of all of the component areas
of biotechnology exceeds the capabilities of one author. Such a book would
also duplicate information already presented in readily available textbooks.
Among these are
- Aiba, S., Humphrey, A., and Millis, N.F. (1973,1976). Biochemical Engineer-
ing. New York, Academic Press.
- Atkinson, B. (1974). Biochemical Reactors. London: Pion, Ltd.
- Atkinson, B., and Mavituna, F. (1983). Biochemical Engineering and Bio-
technology Handbook. New York: Nature Press; London: MacMillan.
- Bailey, 1.E., and Ollis, D.F. (1977). Biochemical Engineering Fundamentals.
New York: McGraw-Hill
- Bergter, F. (1983). Wachstum, von Mikroorganismen~Experimente und
Modelle. lena: G. Fischer.
- Biryukov, V.V. (1985). Optimization of Periodic Processes for Microbial
Synthesis. Moscow. (in russ.)
- Kafarow, W.W., Winarow, A.l., and Gordiejew, L.S. (1979). Lesnaja
Promyshlenost (Modeling of Biochemical Reactors). Moscow. (in russ.)
- Malek, I., and Fencl, Z. (1966). Theoretical and Methodological Basis of
Continuous Culture of Microorganisms. New York: Academic Press.
- Moo-Young, M. (ed.-in-chief) (1985). Comprehensive Biotechnology, Vol.
1-4. Oxford: Pergamon Press.
- Moser, A. (1981) BioprozeBtechnik, Springer-Verlag, Vienna, New York.
- Pirt, S.1. (1975). Principles of Microbe and Cell Cultivation. Oxford: Blackwell
Scientific.
- Prave, P., Faust, U., Sittig, W., and Sukatsch, D.A. (1982). Handbuch der
Biotechnologie. Wiesbaden: Akademische Verlagsgesellschaft.
- Rehm, H.l., and Reed, G. (eds.) (1982ff). Biotechnology~A Comprehensive
Treatise, Vol. 1-8, Deerfield Beach, Fl. and Basel: Verlag Chemie Weinheim.
- Schiigerl, K. (1985). Bioreaktionstechnik. Vol. 1. Aarau Salle u Sauerlander.
Figure 1.3 schematically illustrates the procedures and different fields that
contribute to process development from concept to technological maturity.
The important component parts (in circles in the figure) are
1. The biochemical-biological basis for the preparation ofthe catalyst (isolation,
culture of pure cell lines, maintenance of strains, enzyme preparation), all
the problems of raw materials (the composition of the culture media), and all
the problems of analysis (chemical analysis; biochemical, microbiological,
and physical methods).
2. The stage of "upstream" processing dealing with the sterilization of raw
materials, with cell culture, and with equipment and instrumentation for
/
/
DESIGN
FIGURE 1.3. Areas of biotechnology potentially contributing to the development of

a process. (From Moser, 1977.)
process control including techniques for carrying out the process on an

experimental scale (measurement and control).
3. Bioprocess technology and its working principles used for planning, eval-
uating, and calculating (modeling) bioprocesses (kinetics and reactors).
4. "Downstream" processing, which includes methods for the separation of
cells, cell rupture, product isolation, and purification.
5. Costs and other economic calculations.
6. Genetics, used for the development of biocatalysts with new characteristics
through mutation, selection, and adaptation.
Understanding of the extent of the problems involved in biotechnology
can best be reached through examining the professionalliter'ature, found in
various journals and review series. The most important of these are
Journals
Acta Biotechnologica
Applied Microbiology and Biotechnology
Biochemical Engineering Journal
Bioengineering News
Biofuture
Bioprocess Engineering
Biosensors
Biotech News
Biotechnologie in Nederland
8 1. Introduction
Bio/Technology
Biotechnology Abstracts (Derwent)
Biotechnology and Applied Biochemistry
Biotechnology and Bioengineering
Biotechnology Laboratory
Biotechnology Letters
Biotechnology News
Biotechnology Newswatch
Biotechnology Progress
Chemical and Biochemical Engineering Quarterly
Chemical Engineering Journal
Chemie-Ingenieur-Technik
Current Biotechnology Abstracts
Enzyme and Microbial Technology
European Journal of Applied Microbiology and Biotechnology
Forum Biotechnologie
Journal of Applied Chemistry and Biotechnology
Journal of Biotechnology
Journal of Chemical Technology and Biotechnology
Journal of Fermentation Technology
Journal of Industrial Microbiology
Pascal Biotechnology-Bulletin Signaletique
Practical Biotechnology
Process Biochemistry
Swiss Biotech
Trends in Biotechnology
Review Series
Advances in Biochemical Engineering
Advances in Biotechnological Processes
Annual Reports on Fermentation Processes
Biotechnology Advances
Biotechnology and Bioengineering Symposia
Developments in Industrial Microbiology
Methods in Microbiology
Progress in Industrial Microbiology
The process orientation ofthis book is illustrated by a comparison between
chemical processes and bioprocesses such as simple fermentation or waste
water treatment as a case of complex fermentation process from a process
engineering viewpoint (Table 1.1). The advantages and disadvantages of the
two processes are based on the characteristics of the catalysts. In comparison
with the usual chemical catalysts, enzyme catalysts are both highly active and
highly selective. In the last decades, even more active chemical catalysts have
been developed; however, this has been costly. Since isolation and purification
of enzymes also have costs, enzymes in pure form are also not inexpensive.
TABLE 1.1. Comparison ofbioprocessing and chemical processing.

Criterion of
comparison Chemical process Bioprocess
Catalyst More or less active and selective Enzymes: highly active and
Expensive if selective selective
Expensive regeneration Regeneration easy due to
microbial growth
"Construction" of new catalysts
with genetic engineering
Reaction conditions Mostly high temperature and Mainly - 25C and - 1 bar
high pressure pressure
Biological demands of nutrients,
biological optimum (cf. Fig. 5.1),
maintenance, mutation
Raw materials Pure if poorly selective catalyst Impure, inactive, and diluted
("unconventional" raw
materials: cellulose, starch, oil)
Process Often multistage processing with One-stage possible without
recovery of intermediates, intermediary product recovery
resulting in low yields (e.g., steroid transformation)
Environmental crisis Potentially better for environment
Fast reaction rates at high Mainly slow reaction rates
concentration level and high Low concentrations
yields High demand in sophisticated
apparatus/equipment,
supplementary education
"Demon of nature" (biological
material, infections, mutations)
Today, use of enzymes in the whole cell as the catalysts is preferred; in general,
fermentation technology is still the least costly option. In addition, catalyst
regeneration is easily attained through cell growth and division.
All of the facts noted in Table 1.1 result from the nature of the catalyst.
A quantitative treatment of the kinetics of biochemical-biological systems,
given hereinafter, will mirror these process engineering characteristics. Thus,
biological process kinetics will have fundamental formulas for processes of
growth and product accumulation which operate with a typically biological
optimum in a relatively narrow range of concentrations, pH and temperature.
The formulas must, of course, also take into account the metabolism, the
effects of the interaction between organism and environment, and the nutri-
tional requirements of the cell.
The process-engineering comparison between simple fermentation and a
complex bioprocess such as that used for waste water treatment, shown in
detail in Table 1.2, brings out the problems involved in quantifying practices
used in complex cases: multiple substrate kinetics operating in either sequential
or parallel form; mixed populations; dependence on pH and 'temperature; the
influence of homogeneous or heterogeneous reactor operation in discontin-
10 1. Introduction
TABLE 1.2. Process engineering comparison between simple fermentation and

complex fermentation bioprocess (e.g., biological waste water treatment).
Complex
Criterion of Simple fermentation fermentation bioprocessing
comparison (lab scale) (technical scale)
Catalyst (population) Mainly pure cultures Mixed population "biocoenosis"
Raw materials Synthetic media with S Complex media with unknown
limitations known and even toxic components
(optimized media)
Analytical methods Expensive, modern (e.g., Cheap and simple, often global
enzymatic analyses, (e.g., chemical, biological oxy-
research oriented) gen demand); need for stability
(long time) and reproducibility
Temperature, pH Constant Gradients
Reactors, mode of Mainly homogeneous type Heterogeneous types dominate
operation (e.g., stirred tanks) in batch or (gradients in homogeneous
continuous mode of operation reactors, transport limitation
Pseudo homogeneous approach in three-phase systems, biofilm
preferred reactors)
Batch, semicontinuous operation
Inflow Constant Changes
Kinetics Idealized conditions: I-S Multi-S limitation
limitation Microbial interactions (multi-X)
Transient operation (dynamics)
Interactions between biology and
physics
Structured approach needed
(for biological cell as well as
for reactors)
From A. Moser, 1981.
uous, semidiscontinuous, or fully continuous operations; unstable behavior

with respect to changes in the quantity or quality of input materials, and
special effects such as multiphase operation (heterogeneities) and the problems
of applying global methods of measurement that reflect only overall kinetics.
Identification of the separate tasks in the area of bioprocess technology
clearly shows that in every case the reaction is closely coupled to the reactor.
The quantification of processes involves the kinetics, the transport phenomena,
and the interactions between them. This fact influences the whole way of
thinking and the whole mode of working. Figure 1.4 illustrates a general
strategy for bioprocess kinetic analysis on the basis of this mode of thinking-
an "integration strategy" (A. Moser, 1982)-which will be outlined in more
detail in Sect. 2.3.1. At the same time, the scheme shown in Fig. 1.4 can serve
as a guide to this book: The sequential steps represent the concepts that will
be developed in the following chapters.
The strategy followed in the book is in agreement with the theory of
deduction (Popper, 1972, 1976) which recently was stressed again for economic-
FIGURE 1.4. Flow chart sheet of the strategy of bioprocess kinetic analysis for different
process situations: stationary/instationary, homogeneous/heterogeneous, differential!
integral, and true dynamic/balanced (frozen) reactor operation. rds, rate-determining
step; qss, quasi-steady-state, (From Moser, A. 1983.)
12 1. Introduction
ecological evaluation (F. Moser, 1981, 1983). This scientific strategy is most
helpful to analyze and synthetize the complex network of integrated phenom-
ena of biology and physics. As integration means more than summing-up, the
success of engineering design work depends on the elucidation of interactions.
BIBLIOGRAPHY
Dechema (1974). Biotechnologie, Studie uber Forschung und Entwicklung. Bonn:
Bundesministerium fiir Forschung und Technologie.
FAST (1980) Sub-programme Bio-society, Research activities Commission of the
European Communities EUR 7105. Brussels (M. Cantley)
Hamer, G. (1985). Trends Biotechnol., 3, 73.
Hines, D.A. (1980). Enzyme Microb. Technol., 2, 327-329.
Moser, A. (1977). Habilitationsschrift, Technical University, Graz, Austria.
Moser, A. (1978). Erniihrung/Nutrit., 2, 505.
Moser, A. (1981). Bioprozesstechnik. Vienna and New York: Springer-Verlag.
Moser, A. (1982). Biotechnol. Lett., 4, 73.
Moser, F. (1981). Paper presented at 2nd World Conference of Chemical Engineering,
4-9 Oct., Montreal.
Moser, A. (1988). Trends in Biotechnology, Vol. 6, No.8, August.
Moser, F. (1983). Chern. Eng. (Lond), August/September, 13.
Popper, K.R. (1972). The Logic of Scientific Discovery. London: Hutchinson.
Popper, K.R. (1976). Die Logik der Forschung. 6th ed. Mohr Studienausgabe.
Rehm, H.J., and Reed, G. (eds.) (1982ff). Biotechnology-A Comprehensive Treatise.
Deerfield Beach, Fla. and Basel Vol. 1-8 Verlag Chemie Weinheim.
Reuss, M. (1977). Fort. Verfahrenstechnik, 15F, 549-566.
Ringpfeil, M. (1977). Chern. Tech., 29, 424-428.
Moo-Young, M. (ed.-in-chief) (1985). Comprehensive Biotechnology, Vol. 1-4, Oxford:
Pergamon Press.
CHAPTER 2
The Principles of Bioprocess
Technology
2.1 Empirical Pragmatic Process Development

Chapter 1 mentioned the central importance of the catalyst in biotechnolog-
ical processes. A culture suitable for production purposes-that is, a cell
culture with sufficiently high population density and capacity for economic
production of the product-must be available.
2.1.1 PRODUCTION STRAINS
Only about 1/10th of all the types of microorganisms present in nature are
known. In principle, a production strain can be isolated from natural sources
or can be constructed by genetic manipulations. Experience has shown that
the potential of a strain to produce a product can be increased by a stepwise
progression from the laboratory scale to the technical scale through screening
stages (Aiba, Humphrey, and Millis, 1976). The complete stepwise screening
program is shown schematically in Fig. 2.1. The steps for preparing the
inoculum for a production-scale reactor begin most often with flasks, and the
volume is increased in steps of about 1: 10 or more until the size of the
production plant is reached (Metz, 1975; Rehm, 1980). The point is to ensure
that the large production reactor will have, from the very beginning, a suffi-
ciently high cell concentration to minimize the danger of infection from foreign
strains and to permit reaching the final cell density desired for production
purposes (normally ca. 2-50 gil) in an acceptably short period of time.
2.1.2 THE STARTING POINTS
In general, four situations are conceivable in starting a program of process

development. In practice, a combination of all four usually exists.
1. Establishment of the Catalyst. Chemical and enzyme catalysts offer alter-
natives to biological fermentation (cf. Table 1.1). Enzyme technology offers
the advantages of specificity and selectivity: The multiple-sub strate-media
14 2. The Principles of Bioprocess Technology
SCREENING PROGRAM
STOCK CULTURE
1ST SCREENING
(PLATING & CUP METHOD)
2ND SCREENING
( SHAKEN FLASKS)
3RD SCREENING
( PILOT PLANT
PRODUCTION BIOREACTOR
FIGURE 2.1. Procedure for transferring the biocatalytic characteristics of micro-

organisms from the laboratory to an industrial-scale operation. (Adapted from Aiba,
Humphrey and Millis, 1976.)
processes of fermentation technology is replaced with a one-component solu-

tion. At the same time, downstream processing, which is normally labor inten-
sive and costly, can be simplified. For a more detailed study of enzyme
technology the reader is referred to the literature (e.g., Rehm and Reed, 1982,
Vol. 7; Wingard, 1972; Wingard, Katchalski-Katzir, and Goldstein, 1976;
Zaborsky, 1973).
2. Work with New Substrates. In the future, one expects that the basic feature
2.1 Empirical Pragmatic Process Development 15
of enzymes (i.e., metabolizing of almost any substrate) may contribute to solv-

ing problems involving the supply ofraw materials, the environment, and the
energy supply. When these biotechnological processes might gain a foothold
in industry depends on economic, ecological, and political considerations.
3. Manufacture of New Products. Information on the impressive possibilities
for products and the large array of them already being produced may be found
in almost every introductory level publication on biotechnology (e.g., Bogen,
1976; Moo-Young, 1985; Rehm, 1980; Rehm and Reed, 1982ff).
4. Application of New Types of Process Techniques. As will be further explored
in Chap. 3, a great number of mixing and aeration systems have been devel-
oped that could be of industrial interest. All of these reactors can be operated
in discontinuous, semicontinuous, or fully continuous modes. The semi-
continuous mode of operation offers some advantages (Pickett, Topiwala, and
Bazin, 1979). However, a fully continuous operation seems to be limited
(infections, strain stability, continuous flow).
2.1.3 DIFFERENT MODES (OR STRATEGIES) IN PROCESS

DEVELOPMENT
In the development of a process, precise information from the basic sciences
is joined with information from the applied engineering sciences. From the
standpoint of process technology, the problem area of every technology is the
need to increase the scale of measurements by which the kinetic phenomena
are coupled with transport phenomena. Findings about the behavior of sys-
tems of cells interacting with their environment, studied in the microbiological
laboratory using 100-ml to 1-1 Erlenmeyer flasks, must be extrapolated to
large plants that might work with volumes in excess of 100 m 3
The principal routes available in developing a process (Moser, 1977a) can
be summarized as follows:
1. Trial and error. This has been successful in the past but requires a long time
and entails much loss.
2. Pure theoretical route via the solution of all differential equations. This
seems to be a utopian path for the future.
3. Engineering approach based on quantification, including pilot plant work
and work with or without mathematical models. Pilot plants are generally
10 times larger than the laboratory reactors but are smaller than produc-
tion-scale plants. In the first step, the volume of the pilot plant reactor is
typically about 50 to 500 1; a second step extends this to 500 to 5000 1.
2.1.4 PROCESS DEVELOPMENT WITHOUT MATHEMATICAL

MODELS
The individual steps of a process development procedure not involving a
mathematical model are shown in Fig. 2.2. The important characteristic
FIGURE 2.2. Process develop-

ment involving no mathematical
models: the empirical pragmatic
approach. X, biomass; Sj, sub-
strate, Pj' products; P/V, power/
volume; Hv TR, volumetric heat
r - ECONor1Y - - - _J rate.
I
I
I
I
I
I
feature of this route is that the type of reactor to be employed is determined

pragmatically or intuitively on the basis of information from the literature or
from experiments carried out according to more or less accidental considera-
tions, such as availability.
The first phase, exploratory research, takes place mainly in the microbio-
logical, biochemical, and chemical laboratories and yields mainly qualita-
tive results concerning the reaction components-the cells and the nutrient
medium.Typical results from experiments in Erlenmeyer flasks include
- Characterization of the cell culture in terms of, for example, the shape of the
cells and the population (morphology).
- Characterization of the nutrient media with regard to temperature, pH,
aerobic or anaerobic characteristics, and composition of a "minimal me-
dium" (the limiting substrate concentration).
- Type of metabolism and range of products (physiology).
Optimization of the media can be done in two ways: (a) balancing the
nutrient solution, the cells, and the remaining fluid (Dostalek et at, 1972;
Haggstrom, 1977; Pirt, 1974), or (b) experiments in a chemos tat to determine
the influence on growth of different concentrations of the components of the
growth medium (Kuhn, Friedrich, and Fiechter, 1979; Mateles and Batat,
1974; Tsuchiya, Nishio, and Nagai, 1980).
After a preliminary estimate of costs to see whether production is economi-
2.1 Empirical Pragmatic Process Development 17
cally justifiable, that is, whether production costs are low enough that the
product can be profitably marketed (see Sect. 2.2.4), a pilot plant is built.
Because of the desire for more rapid development and minimized costs, the
pilot plant is usually constructed on the smallest possible scale. The pilot plant
gives results concerning "space-time-yield" relationships (see Sect. 2.2.3) and
some economic factors (Hv TR, P IV), and provides indications about product
quality and the efforts required in the final phases of separating and purifying
the product. The pilot plant also aids in selecting the criteria for a scale-up
(Hockenhull, 1975; Oosterhuis, 1984). This choice, like the choice of reactor
type, results most often from pragmatic considerations and experience with
the theory of similarity. In the case of presumed geometric similarity, the
following quantities are used as pragmatic scale-up criteria, if a systematic
scale-up is executed at all:
- P [W], resp. PGIV [kWm- 3 ]
- Fp , pumping capacity of stirrer, resp. FL [m 3 . S-I] or FGIVL
- Vtjp [m S-1 ] or y [S-I]
- Rest [ - ]
-k L a[h-l]
- ro[kgO z m- 3 h- 1 ]
- VS L [m 3 m- 2 s- 1 ]
- t m L or t c L [s]
- Deff L [m z . S-I] in case of continuous plug flow reactor (CPFR)
All of these criteria are dependent variables. The variables available for a
theoretical calculation are VL , Pu d j , n, dT , and FG
At the moment there is no way to attain a proper scale-up, although in the
recent past there have been important developments in this field (Kossen and
Oosterhuis, 1984; Moser 1977a; Ovaskainen, Lundell, and Laiho, 1976; Reuss
et aI., 1980).
With this type of process development, it is more interesting to establish the
special conditions of an essentially given mode of operation in an also given
reactor than to attempt optimization. The construction of a mathematical
model, and any later systematic optimization of processes in the individual
reaction steps, are in practice not done. The time and expense associated with
model building are often thought to be economically unjustified: Quantifi-
cation ofthe experimentally measured conversion rate, or productivity (yield),
is sufficient, without separation of phenomena into biological and physical
components. In practice, the space-time-yield mode of thought will continue
to be justified in most cases.
A more systematic process development strategy extends beyond simple
quantification of yield based on gross quantities (see Sect. 2.3.2). It involves
mathematical models that are, among other things, useful for later optimiza-
tion or computer-coupled systems operation (Stephanopoulos and San, 1984).
In the future, models will be used together with systematic process design
whenever (a) competition with other industries or the crisis in the environ-
ment, in raw materials supply, or in energy supply make optimization of the
process and (b) improved methods of data acquisition, handling, and evalua-
tion are available for key variables.
2.2 Basics of Quantification Methods for Bioprocesses

Figure 2.3, a typical schematical diagram for a bioprocess, represents the
activities of microbial cells according to the macroscopic principle (see Sect.
2.4.1). In most cases the organisms live in the liquid phase (L) and the nutrients
Sj and gases (such as O 2 in the case of an aerobic process) must be transported
through phases and interfacial surfaces. The products Pj ' CO 2 as a gas and
heat (Hv = volumetric heat of reaction, J .1- 1 ) must be transported away. The
possible limiting steps involving transport are shown in Fig. 2.3 along with
the steps numbered 1 to 4a, which will be important later (see Sect. 4.2).
Quantitative treatment is possible only for variables that are measurable.
Such variables should be easily, reliably, and rapidly measurable in practice.
Further, the variables measured should be key variables. The future develop-
ment of measuring instrumentation will certainly bring further progress in this
area. In practice, data collection is still dominated by the process variables
listed in Table 2.1.
The quantitative treatment of such a simplified process scheme with six
variables is derivable from the following gross equation with stoichiometric
THE f'lAC ROSCOP IC PRINC IPLE WIT H CELLS AS "BLACK BOXES"
s.I
BALANCE -LWE
(REACTOR)
FIGURE 2.3. The macroscopic principle applied to bioprocessing expressed with

pseudohomogeneous observable process variables in the liquid phase (L): biomass X,
substrates Sj , oxygen 0, products Pj ' carbon dioxide C, and volumetric heat Hv.
Pseudohomogeneity is checked by considering a series of mass transfer steps: L film
at the gas phase-L interface (1), L bulk (2), L film at the L- Solid phase (S) interface
(3), cell wall and membranes (4), resp. S-phase cell mass with cytoplasm.
2.2 Basics of Quantification Methods for Bioprocesses 19
TABLE 2.1. Conventional process variables and their measurements.

Symbol Process variable Measuring method
x Biomass Cell dry weight. turbidity. cell number
S Substrates Enzymatic analyses-global methods (e.g. biological.
chemical oxygen demand)
Products Enzymatic analyses or special method
Oxygen p O 2 -electrode, gas analysis
Carbon dioxide Pco, -electrode, gas analysis
Heat of fermentation Temperature, heat balance
coefficients Vi:
VsSj + V002 + Xo ~ VxX + VPPj + VCC02 + vHHy (2.1)

This equation holds in general for discontinuous fermentations operating
aerobically with the production of CO 2 , In some cases, of course, such as
reactions with algae, CO 2 is used and O 2 is produced. In principle, all fermen-
tation processes are autocatalytic over a wide range-the presence of cells
promotes the reaction. This is expressed in Equ. 2.7.
A quantitative treatment of bioprocesses includes the following basic
concepts:
- Stoichiometry and thermodynamics
- Rate of the biological process
- Productivity, conversion, yield
- Costs, profits
2.2.1 CONCEPTS OF A UNIFORM NOMENCLATURE FOR

BIOPROCESS KINETICS
1. There is a distinction among symbols for the absolute rate of a reaction (r;),
the rate of transport of a substance (n;), and the rate of heat (q). Lower case
italic letters are used; the dot indicating the first derivative is omitted.
2. The specific, or relative, rates of growth (u) and of synthesis (qp or qe or qHJ
and consumption (qs or qo) are distinguished from the absolute rates of
consumption (rs or ro) or formation (rx' rp ' re , rHJ
3. Rate constants are always designated with k, uniformly for both reactions
and transport: k or kr or kTR with k L , k s , kG, and k H ,. The rate of reaction,
with n the order of the reaction is generally
(2.2a)
The dimensions of the reaction rate constant, k" are [t-l (moljV)l-n] thus,
dependent on the order of the reaction. For the rate of mass transfer (with
Ac = the concentration gradient):
(2.2b)
TABLE 2.2. Basic concept of unified nomenclature in bioprocess engineering.

Variable Biokinetics M ass transport Heat transport
Rates dC i
dC i dhv
Absolute (e.g., for dis- ri = Tt ni = Tt q=Tt
continuous stirred tank
[g.!-Ih- 1 ] [g.\-Ih- 1 ] [kH-l'h- 1 ]
reactor)
rx, rs, ro, rp, fe, rHv
dC i .~
Relative (specific)
dt x
p, qs, qQ, qp. qeo qH.
Rate constants (coefficients) k, kT'
r,=k'(c)n k p , k d ... ku, k L2 ks,
ktot , kg
Stoichiometry
Microscopic: stoichiometric
coefficients
VA' a -> VB' b + vc' c Vji(V A , VR,')
Macroscopic: yield Yjli
coefficients
1
1',= -y 'rj
jIi
Equilibrium constants
and for the heat transfer rate (~T = temperature gradient):

q = kHTR'~T (2.2c)
Here, kTR and kHTR are phenomenological transfer coefficients, which are
further interpreted using true transfer coefficients (kLl' kLl' k H) and the
corresponding interfacial areas (aGIL' aLls, aH)'
4. The symbol Y is proposed for yield coefficient (see Equ. 2.13).
5. The stoichiometric coefficients are designated as Vi' according to Equ. 2.1.
6. The concentrations of substances are indicated with lowercase letters, in
accord with the nomenclature of the enzyme literature. Capital letters are
used for quantities of materials (e.g., X = x' V).
7. The symbol K is exclusively used as the equilibrium, saturation, or inhibi-
tion constant in a kinetic equation.
Tables 2.2 and 2.3 summarize the nomenclature used in this book and
compare it with the nomenclature used elsewhere.
2.2.2 THE RATES OF A BIOPROCESS

The concentration-time profile for the six process variables shown sche-
matically in Fig. 2.3 is shown in Fig. 2.4. This is for the case of a discontinuous
process with constant reactor volume YR' Conditions that can be observed and
TABLE 2.3. Symbols used for rates in bioprocessing. IV
N
specific rates [h -I] tXl
!>O
this textbook: '"(').
concentration Absolute rate Pirt Moser IUPAC
Varible (g. I-I) (g'I-"h- ' ) positive negative (1975) (1981) proposal (1982) '"0-.
tJ
Enzyme, E e rE k cat ~
!>O
Cell mass, X x rx 11 (Ild' k d, k.) 11 11 11 1:1
....
Cell number, N n rN v v 5i
(')
Substrate, S rs qs (ms) qs(ms) 0"(0".) qs !>O
....
Oxygen, 0 0 ro qo (mo) qo, 0"0(0"0 ) qo, o
1:1
[mmol/gX' h]
~
('p
Product, P p rp qp qp n qp ....
Carbon dioxide, C 1:1"
c rc qc (mcJ qco, ne qeo, 0
0-
Heat, Hy hy [kH- ' ] rHo [kJ"-"h- ' ] qH. (mH.l n H
'"0'
...
tXl
o
"0
...0
(')
('p
('p
'"'"
'"
IV
......
-t
FIGURE 2.4. Typical concentration-time curves for fermentation process in discon-

tinuous culture, with reactor volume constant. S, substrate; 0, oxygen; X, biomass; C,
carbon dioxide; P, product; H" heat.
measured are growth of biomass X, consumption of Si and O 2 , and accumula-

tion of Pi' CO 2 , and Ifv. The rates involved are of prime interest for reactor
balancing and kinetics. In formulating the rate equations, one must derive or
define them from the conservation of mass equation (see Sect. 2.3.1). This
equation contains two principal terms: (a) sources and sinks, where compo-
nent i is made or used, and (b) the mass flux, n; [g/cm 2 s], of component i
through the surface volume element under consideration.
The flow of mass n; = deN / A)/dt can be due to various kinds of transport
phenomena, and it is described by various terms (for example, the convective
flow velocity v; or conduction through a gradient, that is, diffusion, D; or
dispersion, Dcff ; or interfacial transport, kTR (see Fig. 2.3).
A change in the concentration Ci with time is expressed in Equ. 2.3. In
general, the equation can be written using a vector operator V (grad = partial
derivative with respect to a direction coordinate) as follows:
OC
-ati = -'In. + r
I
I - I
(2.3a)
Separating the various transport terms with n; = Cj v and v = F / A, this equa-

tion can be written in vector notation as
Oc.
~ = [
vi
-div(c j v) + diveD grad cj ) + kTR l1cJ + r
- i
(2.3b)
or, in one-dimensional form (for example, for tubular reactors, cf. Equ. 3.4), as
(2.3c)
The rates of formation and/or consumption of component i are intensive

variables [g. 1-1 . h -1 ], and they differ from each other in that the formation
rate is positive and the consumption rate is negative. Equation 2.3c can be
reduced to Equ. 2.3d for the rate of formation or consumption in a discon-
tinuous stirred tank reactor:
(2.3d)
Historically, this was the first rate equation, and later it was arbitrarily used
as the definition of a "true" reaction rate. However, Equ. 2.3d is valid only with
VR = constant, T = constant, and ideal mixing. Furthermore, it represents
only the rate of formation or consumption of a component. Other cases, with
other reactor configurations, will be discussed in Chapter 3; in that Chapter,
due to the introduction of a transport term, more complex formulas define
rates.
We can now express the rates associated with the model process of Fig. 2.4
for the case of an ideally mixed discontinuous reactor with VR = constant and
T = constant. The rate of growth (biomass formation) is
dx
rx =Tt (2.4a)
The rates of oxygen and substrate utilization are

ds do
r --~ and (2.4b)
s- dt dt
The rates of synthesis for P, C, and Hv are
dp de
r-- r-- (2Ac)
P - dt' C - dt'
These rates have dimensions of [g .1- 1 . h -1] or [kJ .1- 1 . h -1] and are intensive
quantities that may be substituted directly into the mass balance equation.
On the other hand, these absolute rates may take on any value and are
therefore not characteristic of the system. Comparable variables, which are
biologically representative, are the so-called specific rates of production or
utilization, which refer to the catalytically active mass (to a first approxima-
tion, the concentration of biomass, x, is taken as the dry cell weight). With
this, one has the definitions ofthe specific rates for bioprocesses of Equs. 2.5a - f
where, in each case, the specific rate has the dimension [h- 1 ]. For growth, the
specific rate is f1
1 dx
_ . - = f1 [h- 1 ] (2.5a)
x dt
or
1 dn
_.- = v [h- 1 ] (2.5b)
n dt
For the consumption of substrate, the specific rate is qs
(2.Sc)
or for O 2
1 do -1
qo = -~. dt [h ] (2.Sd)
The specific rate of product accumulation is qp
(2.Se)
and for CO 2
(2.Sf)
and for By
(2.Sg)
These definitions for specific rates are analogous to the definition of specific
rates rt in chemical kinetics, where the mass of the active catalyst, Me in the
case of heterogeneous catalysis, is used to define a specific rate
1 dN
r.*=--' (2.Sh)
, Me dt
The reactor volume VR(V) is used predominantly in the case of homogeneous
catalysis
1 dN
r.*=--' (2.Si)
, V dt
In some cases other quantities are useful, for example, the external catalyst
surface or the volume of the solid phase.
The quantity Nj in Equs. 2.Sh and 2.Si represents the number of moles of
component i (c j = N)V). The connection between the definitions of Equ. 2.Si
and Equ. 2.4 can be made by substituting this fact
rt = V1 . dtd(c j ' V) =
dcdt' + (c.;. dV)
dt (2.Sk)
Equation 2.Sk is identical to Equ. 2.3d for the case of V = constant.

For the sake of completeness, Equ. 2.Si is often taken as the definition of
the reaction rate in chemical kinetics. Nevertheless, the true reaction rate, r,
is connected with the previously indicated and used rates of growth or con-
sumption, rj or rt, through the stoichiometric coefficients Vj
r
r = -i (2.6)
vi
For consumption, Vi < 0 and for growth, Vi > O. Naturally, a component can
also have no reaction rate. A reaction rate can only be defined in cases where
both consumption and formation occur as indicated by Equ. 2.6.
For bioprocesses, the stoichiometric coefficients are seldom known (Cooney,
Wang, and Wang, 1977), so the reaction rate in the strict sense of Equ. 2.6 can
seldom be used. However, rates of consumption or formation of the type given
in Equs. 2.5a-g are well defined.
Finally, the autocatalytic nature offermentation becomes evident from Equ.
2.7, which can be derived from Equ. 2.5a
rx = fl' x (2.7)
2.2.3 STOICHIOMETRY AND THERMODYNAMICS

2.2.3.1 General Remarks
Considered as the study of the quantitative composition of chemical com-
pounds and the quantitative conversion of chemical reactions, stoichiometry
belongs to the fundamentals of reaction engineering, along with thermo-
dynamics and kinetics. Industrial plants for bioprocessing on a large scale will
be expensive to construct, and therefore maximum profit will have to be
extracted from the process. Intensive studies of the quantification of metabolic
reactions in different reactor configurations are required, studies that examine
all factors making for maximum yield. Hence the principle of balancing plays
a central role. One can distinguish between macro balances and micro balances.
For conservative properties, which are not altered during the reaction, only
macro balances are used; for nonconservative properties, microbalances are
used as well (Kossen, 1979).
In contrast to kinetics, where laws are useful for predictions, in stoichio-
metry calculations have to be based on knowledge of the reaction mechanism,
the balances of the initial values of the significant compounds ("key variables").
It must be remembered that, just as in the balancing of chemical processes,
macroconversion in bioprocessing is always the sum of transport phenomena
and biological transformation reactions. As a consequence, Equ. 2.3 must be
applied in all special cases of reactor operation modes (see Fig. 2.5). Modifi-
cations are needed for complex reactions and complex reactor operations.
In this basic equation, all facets of reaction engineering sciences are manifest:
balancing, stoichiometry, kinetics of reaction rates, and physical transport.
Thermodynamics are not directly included here because thermodynamics
emphasizes only initial and final states of the system, whereas kinetics deals
primarily with the time required for the transitions involved. Stoichiometry
includes the rules linking changes in the composition of reaction components
during the process. Thermodynamics gives information about the maximum
Stoichiometry
~Complex
I
SimPle--------
Comp'" ~, \
Simple reactions in
closed (discontinuous stirred tank)
J Complex reactor configurations
Semidiscontinuous. semicontinuous operation
or steady-state open (continuous Non-stationary techniques
stirred tank or plug flow) reactor
systems
?
Mole balances
Problem of "stoichiometrical dependence"
Search of "key variables" and "key reactions"
Yield
Conversion
Extent of reaction
FIGURE 2.5. Overview of the problems involved in stoichiometry oftechnical processes.
possible extent of reaction and the heat liberated or absorbed and also permits
calculation of the equilibrium constant from the standard free energy.
The activities of microbial cells in bioreactors can also be expressed by the
balance equation for the chemical elements, instead of Equ. 2.1, considering
the conservation of the atomic species (carbon C, hydrogen H, Oxygen 0,
nitrogen N):
VS(CaHbOcNd) + VO (02) + vN (NH 3 )
--+ vx(CaHpOyNo) + vp(CaHp,Oy,No') + vc(C0 2 ) + vw (H 2 0) + vH ,' Hv
Here, CaHbOcNd is a generalized carbon source, CaHpOyNo a cell unit, and

Ca, H p, Oy, No, a product. In special cases, the elements sulfur and phosphorus
can also be included.
The overall biosynthetic equation, Equ. 2.8, is the net equation resulting
from hundreds of metabolic reactions in the living cell. The various cycles and
chains in the metabolic network, including the ATP system and other energy-
handling systems that do not result in the output of new cells or products, do
not contribute to the net reaction. Hence, detailed knowledge of these cycles
is unnecessary in a macroscopic treatment using the net stoichometric equa-
tions. Bioprocess analysis, therefore, is made less complex with no significant
loss of information.
The general form for the conservation of atomic species in the case of
multiple reactions, expressed as a sum, is given by
(2.9)
Direct application of Equ. 2.9 to the stoichiometry of bioprocesses does,

however, present some difficulty, because several hundred components take
part in metabolism. This problem has thus far been avoided, and only a limited
number of compounds have been considered in a so-called gross stoichio-
metric equation of growth. This approach has been theoretically justified
(Roels, 1980a).
A practical implication of Equ. 2.3 using Equ. 2.6 in a case of multiple
reactions in the steady state, with ocj/ot = 0, written as a sum
L Vnj = - L vj r (2.10)
j j
is that the stoichiometry of a reaction pattern can be readily studied experi-

mentally in steady state by observations of the exchange flows (Vnj) with the
environment. There is a group of extensive quantities, called conservative
quantities, that have the property that in the reaction pattern of the system
no net production take place (Roels and Kossen, 1978). The balance equation
for conservative properties has no production term
LVnj = 0 (2.11)
j
In the steady-state case, the reasoning presented formally validates the treat-
ment of microbial metabolism in terms of a gross stoichiometric equation
of growth and product formation (Roels, 1980a). This formalism, however,
results in complicated equations that offer little advantage over other equa-
tions. Therefore, only situations oflimited complexity can be studied in detail.
The main problem in applying stoichiometric considerations to bioprocess-
ing (beyond quantification in non-open-reactor systems) arises from the com-
plex metabolic reaction network. In simple reactions stoichiometry is trivial,
and complex reactions can only be handled with the aid of a formal mathe-
matical approach analogous to the approach for complex chemical reactions
(Schubert and Hofmann, 1975). In such a situation, an elementary balance
equation must be set up. Due to complexity, it is not surprising that the
approach first used in the quantification ofbioprocesses was much simpler-
the concept of yield factors Y. This macroscopic parameter Y cannot be
considered a biological constant.
2.2.3.2 Definition of a "Mole" of Microorganisms

Several suggestions have been made as to how the molar concept may be
applied to microorganisms. Elementary analysis, performed by standard Pregl
microcombustion techniques on cells, has shown that
1. There is no unique empirical formula for a microorganism-the elementary
composition varies with the growth rate and limiting nutrient. For yeast,
for example, the following "formula" was given (Harrison, 1967):
or
For aerobic microorganisms, the four principal elements are present in a
ratio ofC s H 7 0 2 N (Hoover and Porges, 1952), which is very similar to that
obtained for anaerobic microorganisms (Speece and McCarty, 1964). A

widely used formula that includes phosphorus is (McCarty, 1970)
(2.12b)
2. There is, consequently, no unique elementary balance equation for micro-
bial growth; rather, there is a range of such equations with different
numerical coefficients.
All quantities considered (e.g., number of organisms containing 1 g' atom of
total nitrogen or of protein-nitrogen, or of DNA phosphorus) vary with
growth rate and limiting substrate. It would seem logical to derive the concept
of a "mole" of microorganisms from the empirical formula that (neglecting
sulfur, phosphorus, and ash) can be expressed in the general form CaHbOcN d.
Finally, by dividing all the subscripts by a, giving the formula CHb/aOc/aNd/a,
the definition of 1 "e-mole" represents the quantity of microorganisms con-
taining 1 g' atom (12.011 g) of carbon (McLennan et al., 1973). The general use
of this formula for expressing quantities of microorganisms on a molar basis
is strongly recommended (Herbert, 1975).
2.2.3.3 The Concept of Yield Factors (Yield Constants)

A parameter that is of great relevance to the economic evaluation of the
potential of, for example, a carbon source for biomass or product formation
is the macroscopic yield factor ;u. This yield factor was originally defined by
Monod (1942) in mass units by the quotient
Yxs = - ~X = Xt - Xo :::::: _ dxjdt = _rx = dX [ g cells formed] (2.13)

I ~S So - St dSjdt rs dS g substrate used
In Fig. 2.6, a schematic evaluation of the yield constant is shown, the value
for ;u can be taken from the slope of the line, representing, for example, the
data rs versus rx The definition of yield factor used by Monod is of a purely
macroscopic nature and can be applied to all compounds, in analogy to Equ.
2.13, to quantify cell or product yields including heat, by a relation between
r.
I
FIGURE 2.6. Evaluation plot for yield coeffi-

cients: Rate of consumption of production
of component i versus rate of production of
+------------ rj consumption of component j.
TABLE 2.4. Summary of macroscopic yield coefficients in bioprocessing.

Yield
coefficient Reaction or relation
Yilj* between components Definition Equ. no. 2.14a-o
YXIS S->X rs= -(l/Yxls ) rx 2.14a
YXIO O 2 ->X ro = -(1/YXIO ) rx 2.14h
YPjs S->P rs= -(l/YPIS)rp 2.14c
Yqs S->C0 2 rs= -(1/Yqs)rc 2.14d
PIO O 2 ->P ro = -(l/YPjo) rp 2.14e
YH,IS S->Hy rs = -(l/YH,ls) rH, 2.14f
Yqo O 2 ->C02 ro = -(l/Yqo)rc 2.14g
YH,IO O 2 -> Hy ro = -(l/YH,lo) ro 2.14h
YXiP P~X rp = (l/YXIP)rX = (YPjx)rx 2.14i
YXIC C02~X rc = (l/Yxlcl rx 2.14j
YJqH, Hy~X rH, = (l/YXIH,) rx 2.14k
YSlo 02~S ro = (1/110) rs 2.141
Yqp P~C02 rp = (I/Yqp) rc 2.14m
YPjH, Hy~P rH, = (l/YPIH,) rp 2.14n
YqH, Hy~C02 rH, = (l/YcIH,) rc 2.140
* Dimension in the case of mass components: [g component i/g component j); dimension in the
case of heat components: [kJ/g component j].
the amounts consumed and the amounts formed. The same concept of yield
can be applied to quantify the relation between consumed or produced
amounts. The definitions of yield factors are summarized in Table 2.4 and
represented by Equs. 2.14a-o.
In some instances, however, it is more advantageous to express yields in the
units of moles:
[ g cells formed ] _ y;mol

(2.15a)
mole substrate used - XIS
or
[ g cells formed ] _ y;g.atom

(2.15b)
g. atom substrate used - XiS
These units are always used for yields of cells in grams per mole of ATP formed
during growth
Ax
YATP=A (2.16)
atp
This yield factor plays an important role when the molecular energetics of
biochemical pathways are considered. Bauchop and Elsden (1960) found that
YATP was about 10.5, independent of the nature of the organism and the
environment. According to Stouthamer and Bettenhaussen (1973), an equa-
tion can be assumed for the consumption of ATP in a metabolizing cell,
YATP = . rx + mATP ' x (2.17)

YATp,max
that has the same form as the linear equation often used for biomass growth
1
rs = - ' rx
Yxls
+ ms'x (2.18)
where ms and mATP are the maintenance requirement for substrate and ATP.
Although the mathematical structures of Equs. 2.17 and 2.18 are the same,
their status in terms of the system's description is totally different. Equ. 2.18
is a macroscopic balance equation concerning the flow of substrate entering
the microbial system. Equ. 2.17 is a microscopic assumption about the internal
functioning of the regulation of the energy transformation processes in the
cell, describing the fate of the energy carrier ATP (Roels, 1980b). Stouthamer
and Bettenhaussen (1973) pointed out that the effect of maintenance energy
on YATP has been underestimated, and it is one factor responsible for the wide
range of YATP values found (5-32 g dry biomass per mole ATP, depending on
the nature of the carbon). The concept of the "C-mole of biomass" allows
yields to be expressed in molar units. This can be done in two ways
[ g' atom cell C formed] -YS

_ c
(2.19)
mole substrate used
and
[ g' atom cell C formed] c

(2.20)
g . atom substrate C used = Yc
While both may be useful, the latter is usually preferred, because it is indepen-
dent of the number of carbon atoms in the substrate molecule (Herbert, 1975).
Applying similar concepts to the important constants YATP and Yo, these
are now expressed as
g. atom cell C formed] _ y:c
[ (2.21)
mole ATP formed - ATP
and
[ g. atom cell C formed] = Y,c

(2.22)
. g'atom O 2 used 0
The quantity Y<f is of practical importance in the derivation of elementary

balance equations and of theoretical importance in connection with oxidative
phosphorylation (Tempest and Neijssel, 1975).
In addition to cell yields from carbon substrates and oxygen, cell yields from
other nutrients perhaps are best expressed in the units of Equ. 2.20, as for
example
Yrf YIf
Other important yield factors and their significance are reviewed in articles
by Payne (1970), Bell (1972), and Atkinson and Mavituna (1983). Various
proposals for macroscopic efficiency measures have appeared in the literature,
and they are all more or less related. The yield can be based on electrons
initially available in the substrate for transfer to oxygen during combustion
of the substrate:
[ g SUbstrate] = Yaye _ (2.23)

electrons
The concept of "available electrons" was first developed by Mayberry et al.
(1968) and was later applied to biomass by Minkevich and Eroshin (1973).
The number of available electrons is a quantitative measure of the energy
content of growth substrates (Eroshin, 1977).
A second measure proposed by Payne (1970) is the yield of biomass dry
weight per kilocalorie heat of combustion of the substrate, AHs:
g cell dry matter ]
[ (2.24a)
kcal heat of combustion of S = Ykcal
which can be deduced for aerobic fermentations, for example, as
y,mol [ g'mol-1 ]
y; -~ (2.24b)
kcal - AHs kcal. mol- 1
or (Prochazka et aI., 1970)

YXIS
Ykcal = AH y, AH (2.24c)
S - XIS x
where AHx is the heat of combustion of cell mass (kcaljg).

These two quantities of YaYe- and Ykcal are related, because the heat of
combustion of an organic compound, AHs , is known to be proportional to
the electrons available for transfer to oxygen (Minkevich and Eroshin, 1973).
2.2.3.4 Some Thermodynamic Explanations of Yield Factors

The concept of available electrons allows the energetic yield of growth to be
expressed as a fraction ofthe available electrons of the substrate being changed
into biomass. The rest of the available electrons are transferred to oxygen,
and their energy is dissipated. The energy efficiency coefficient of growth 11E
is the part of the available electrons that are transferred to biomass. 1 - 11E is
the energy consumed during the process of biosynthesis and released as heat.
This efficiency measure, 11E' is proportional to YaYe- as well as to Ykcal and
seems to be very promising in applications, although Roels (1980a) has sug-
gested that the correct formulation of the second law of thermodynamics for
open systems is missed.
The various efficiency measures have been reviewed by Nagai (1979) and
are used in studying the energetics of growth and product formation (Erickson
et al. 1978; Roels, 1980a). In the case of, for example, aerobic growth on a
single source of carbon and energy without product formation, the efficiency
factor for oxygen of Minkevich and Eroshin (1973) is shown to be related to
the "degrees of reduction" of substrate Ys and of biomass Yx:
Y~ls' Yx
lJE=--- (2.25)
Ys
with
V' _ Yxls _ Vx
~XIS---- (2.26)
a vs/a
where a is the number of carbon atoms in one molecule of substrate. Equ. 2.26
is a somewhat modified definition of the yield factor in Equ. 2.13, which
is more convenient for the treatment to be presented (Roels, 1980b). Y~IS
expresses the mole dry weight of biomass per carbon equivalent substrate
consumed.
The correlation in Equ. 2.25 is numerically different for different nitrogen
sources because Yx depends on the nitrogen source used (Roels, 1980a). These
generalized degrees of reduction are analogous modifications of the original
definition of the "reductance degree" (Minkevich and Eroshin, 1973) defined
by
Ys = 4a + b - 2c - 3d (2.27a)
and
Yx = 40: + /3 - 2y - 3(5 (2.27b)
The reductance degree of biomass, Yx, is the number of equivalents of O 2
required per quantity of biomass containing 1 g' atom of carbon; and Ys, the
reduction degree of substrate, is the number of equivalents of available elec-
trons in that quantity of organic substrate that contain 1 g' atom carbon,
according to the general balance equation for aerobic microbial growth on a
single carbon and energy source with NH3 as the nitrogen source.
The number of equivalents of available electrons is taken as 4 for carbon
(a, 0:), 1 for hydrogen (b, /3), - 2 for oxygen (c, y), and - 3 for nitrogen (d, (5).
It is interesting to note that a relationship exists between Y~ls and Yave -
(Roels, 1980a)
Y~IS
Yave -- = - ' Mx (2.28)
Ys
where Mx is the molecular weight of the biomass [g/C-mole]. In place of the
efficiency factor lJE' which was based on an incorrect statement of the second
law of thermodynamics, Roels (1980a) developed the concept of the "thermo-
dynamic efficiency" lJth
(2.29)
FIGURE 2.7. Thermodynamic efficiency of 11th ENERGY LIMITATION

aerobic growth 'Ith with one carbon
source as a function of the substrate's 2ND LAW OF
THERMODYNAM I CS
degree of reduction y., and theoretical
limits of the efficiency (Roels, 1980a. With CARBON
LIMITATION
permission of John Wiley & Sons, Inc.). v
0.5
FIGURE 2.8. Relationship between sub-

strate yield factor Y~ls and substrate CARBON LIMITATION
degree of reduction Ys, and the theoretical
limits to Y~IS (Roels, 1980a. With permis-
sion of John Wiley & Sons, Inc.). 0.7 0 0 00
8 ... __ 08
0 0
LIMITATION
8
LAW) ~ ""tJO.__ 0
;8 ~
0.5
I 0 -"
0
I
~ ~
I
I
,J"
ct'
0 )Is
5
where the maximum value of the yield is calculated in full accordance with
the second law. This equation has been used in the analysis of some recent
yield data from the literature. The data are presented graphically in Fig. 2.7:
The thermodynamic efficiency 11th is plotted against the degree of reduction Ys
of the substrate for growth, with NH3 as a nitrogen source. Figure 2.7 shows
appreciable variation; the trend seems to be adequately represented by the
dotted line. The tendency for 11th to be low for highly reduced substrates (e.g.,
methane) becomes clear, and probably it is also low for highly oxidized sub-
strates (e.g., oxalic and formic acids). Due to carbon limitation, Y~ls can never
exceed 1; hence, the value of 11th is subject to the restrictions given by the right
part of the solid line in Fig. 2.7. The left part ofthe solid line is the restriction
due to the second law; 11th can never exceed unity. Figure 2.8 shows the same
results as Fig. 2.7 but plotted in a somewhat different way. The carbon
conversion into biomass Y~IS is plotted against substrate degree of reduction
Ys. Again, the solid line gives the restrictions mentioned.
The tendencies observed in Figs. 2.7 and 2.8 can be summarized as follows:
For substrates up to a degree of reduction of about 4.2, the degree of reduction
of biomass (the energy content of the substrate) is insufficient to allow all sub-
strate carbon to be converted to biomass, even if the thermodynamic efficiency
were unity. For substrates with a degree of reduction higher than 4.2, as far
as energy requirements are concerned, all carbon can be converted into
biomass, and even CO2 could be fixed; the extent to which the energy can be
stored in biomass becomes limiting. Hence, the thermodynamic efficiency
must decrease with growth on substrates of a high degree of reduction. For
unknown reasons, the carbon conservation efficiency Y';'!S never exceeds 0.7.
2.2.3.5 Relationship Between Yield Factors and True Stoichiometric

Coefficients
On the basis of the general stoichiometric equation of microbial growth, given

in Equ. 2.8, the following balance equations for each of the main elements
[C, H, 0, N, neglecting S, P, and ash] can be written
Starting material Products
g'atom C: vS'a = Vx ' rx + Vc + Vp' rx' (2.30a)
g'atom H: vsb + 3vN = Vx ' P + Vp' P' + 2vw (2.30b)
g' atom 0: Vs' C + 2vo = Vx'Y + vp'y' + 2vo + Vw (2.30c)
g'atom N: vS'd + VN = Vx ' c5 + Vp' c5' (2.30d)
In Equs. 2.30a-d, the quantities of substrate (vs) and of NH3 (VN) required to
make 1 C-mole of cells can be measured directly, and the composition of cells
and products can be determined by elementary analysis, giving rx, p, y, c5 and
rx', p', y', c5'. In the case of no nitrogen in the substrate and product, VN = Vx ' c5
and the determination of NH3 is not strictly necessary. However, it provides
a useful check on overall accuracy. O 2 and CO 2 can be determined with gas
analyzers, giving Vo and Vc' The only quantity that cannot be conveniently
measured is vw, the moles of H 2 0 produced. This value, however, can be
obtained by difference from Equs. 2.30b and 2.30c, and the values so obtained
should balance. From Equs. 2.8 and 2.30a-d, a number of important yield
constants can be derived, e.g.,
_ vx{l2rx
YX!S -
+ P+ 16y + 14c5}/(1 - r} [ g cells formed ]
---=-=-'-----~:-'-----,:---'----:---:---"-'-::-------'- (2.31)
vs(12a + b + 16c + 14d} g substrate used
where r is the fraction of cell mass represented by ash (r ~ 0.07-0.1O) (Wang
et ai., 1979). Other yield factors, valid for cases without product formation,
can be expressed as follows (Herbert, 1975)
c 1 [g. atom cell C formed]
(2.32)
Ys = % mole substrate used
yee = _1_ [ g' atom cell C formed ] (2.33)

Vs a g' atoms substrate C used
e Ve [ mole CO2 formed ] (2.34)
Yeo = - -
2 Vs' a g' atom substrate C used
Y~ = VX ' oc = _1_ [g . atom all C formed] (2.35)

2vo 2vo g . atom 02 used
In a similar manner, the yield of product becomes
vp (12oc' + p' + 16y' + 1415')
(2.36)
YPls = vs(12a + b + 16c + 14d) .
In practice, it is usually convenient to calculate the values of these yield
coefficients from experimental data and to use them to obtain the values of
Vs, Vo, and Ve for use in the cell balance equation.
2.2.3.6 Yield Coefficients in Real Fermentation Processes

Yields in real fermentations are not constant, so that the concept of "yield
constants" must be modified. Yield factors exhibit a significant dependence
on various biological and physical parameters. It must be remembered that,
although the yield factor is defined for a given strain with respect to a
particular substance, yield is not only a function of the substrate. Generally,
the following dependence is valid:
Y = f(strain, substrate; Jl, m, S; t, t m , OTR, CfN-ratio, PfO-ratio). (2.37)
Earlier studies on Yx1s (Humphrey, 1970; Wang, 1968) and other reports have
also illustrated the dependence of Yx10 and lical on YxlS (Guenther, 1965;
Mateles, 1971). Pirt (1966) proposed the following function to demonstrate
the relationship of Y to the specific growth rate Jl, as realized in .a CSTR
operating at different dilution rates D (D = Jl), and to the maintenance coeffi-
cient m
1 1 ms
- = max+- (2.38)
Yxis YxlS Jl
or its equivalent form, shown as Equ. 2.18. y~S'x is the value of Yxis as Jl
approaches infinity. Parameter estimation can be carried out, as shown in Fig.
5.27 (cf. Fig. 2.7). Equation 2.18 can be used in modified form to calculate the
influence of Jl and m on Yxis (Abbott and Clamen, 1973).
v _ Jl' YxiS'x
(2.39)
~XIS -
Jl + Y~S'xms
Figure 2.9 shows the general behavior of the dependence of Y on Jl. Beyond
Yxis the values of Ykca1 and YxlO are also shown according to Abbott and
Clamen (1973). Simultaneously, another dependence is shown in Fig. 2.9,
2 0,5
RICICA
o~--+---~--.---.---.---.----r---r---.-------p
o o o 0.2 0.6
FIGURE 2.9. Dependence of yield factors Ykcab YX10 , and Yx1s on specific growth rate /L
"Ricica" refers to data from Escherichia coli with a nitrogen source as the limiting
substrate. (Adapted from Abbott and Clamen, 1973; data for E. coli from Ricica, 1969.
With permission of John Wiley & Sons, Inc.).
2 0,5
Yx/o
YKCAL
01----r---r-----.-----,,-----.-----,------.-- n1 S
o 0 0 0,1 0,2
FIGURE 2.10. Dependence of yield coefficients Ykca1 , YX1o , and YXls on substrate main-
tenance coefficient ms' (Results according to Abbott and Clamen, 1973. With permis-
sion of John Wiley & Sons, Inc.).
derived from data from Escherichia coli with a nitrogen source as limiting
substrate (Ricica, 1969).
The rates at which the yield coefficients change as fJ. changes depend on the
value of the maintenance coefficient Ins as shown in Fig. 2.10.
In a number of cases this approach is not satisfactory, because it looks at
substrate rather than ATP. Therefore, an equation with formal similarity to
Equ. 2.38 has been proposed for the specific production of ATP (cf. Equ. 2.17,
where tJATP is a linear function of f1 (Stouthamer, 1980; Stouthamer and
Bettenhaussen, 1973). More detailed investigations show that, for example,
the maximum specific growth rate is determined by the rate of energy produc-
tion in the cells. As ATP production is correlated with substrate consumption,
the growth behavior can be formally represented at the level of mass, that is,
f1 = f(s).
The same background to formal modeling is evident in the several attempts
that have been made to explain the biomass yield variation with specific
growth rate based on Equ. 2.18 (Reuss et aI., 1974; Rock et aI., 1978). Such an
explanation does not appear to be fully justified. Equation 2.38 cannot explain
why the yield changes during the exponential phase of batch cultures when f1
is constant. Furthermore, it is not true that, in the case of CI-compound
utilizers, Yxls increases monotonically with f1 (or D for steady-state CSTR).
Papoutsakis and Lim (1981) stressed the carbon-flow-branching concept to
illustrate the general possibility of a highly variable biomass yield:
(2.40)
where 1"1 and 1"2 are the rates of branching into metabolic pathways 1 and 2
for the carbon source, Mx and Ms are the molecular weights of the biomass
and substrate, and "x is the stoichiometric coefficient of the reaction S --> X.
In the case ofmethylotrophs, two distinct carbon-flow processes exist: assimi-
lation (1"2) and oxidation (rl)' Cells produce only as much NADH and/or ATP
through oxidation as may be required for assimilation. This fine tuning of the
two processes produces maximal possible biomass yields. According to Equ.
2.40, the yield will change only when 1"1 or r2 changes, and this can happen if
the concentration of any substrate or of any other nutrient or the culture
condition such as T and pH changes. This concept shows that the unusual
behavior of biomass yields of methylotrophs is a kinetic rather than a bio-
synthetic question.
An alternative concept has been proposed by Agrawal et ai. (1982), who
achieved the desired variability of stoichiometry by defining a yield factor that
increases linearly with substrate concentration:
(2.41 )
where a, and b are constants.
The influence of physical parameters on yield coefficients can be quantified
on the basis of the scheme that is valid for yeast metabolism:
~x (2.42)
SI,~,~//. KR P ...... Sz + 0 ~ X.
The "selectivity," (5, of this process with simultaneous glucose and oxygen
effects can be written

Xmax Yx1s
(J = - = - (2.43)
Pm x YPls
Yield, that is, the selectivity, depends on OTR, t m , t e , or CTD. Thus Sac-
charomyces cerevisiae works as a test organism for micromixing if OTR is held
constant (Moser, 1977b).
2.2.3.7 Temperature Dependence of Yield Factors

As a stoichiometric coefficient, the yield factor is not expected to vary signifi-
cantly with temperature (Moletta et aI., 1978; Peters, 1976; Shiloach and
Bauer, 1975; Topiwala and Sinclair, 1971). Temperature seems to have only
a minor influence on Yx1s . In the case of the yield factor YxlO based on BOD,
for example, a decreasing value with increasing temperature is recorded
(Peters, 1976).
2.2.4 PRODUCTIVITY, CONVERSION, AND ECONOMICS (PROFIT)

The productivity, rj , of a process is generally defined as the mass of the
product,j, produced per unit time and per unit volume; it has the dimensions
[kgm- 3 h- 1 ]. To clarify this concept, the curve (Fig. 2.11) demonstrates rj
for the example of the discontinuous process whose course was diagrammed
in Fig. 2.4. With discontinuous processes, a certain lag time is also necessary
between production cycles due to harvesting, emptying, cleaning, and refilling
operations. This lag time, to, is shown on the negative time axis in Fig. 2.11.
- - ------------=---......,,-~
/
/
toO
FIGURE 2.11. Schematic representation of optimal operating point for a process based
on different economic criteria. Productivity may be obtained from the tangent to the
curve. Reaction times: t 1, maximum productivity in a continuous process; t 2 , maximum
productivity in a discontinuous process with to = dead time; t 3 , maximum profit; t 4 ,
minimum cost; t 5 , maximum conversion~maximum product concentration as S -> o.
Drawing a tangent from this point to the concentration/time curve, one

obtains at point 2 the value of the product concentration that can be reached
in the whole production time (ttot = to + t r ). The maximum productivity
attainable with a discontinuous process can be calculated from
Cmax J' Co J'

r - ' -
, (2.44)
j,de.max - to - tr
Point 1 in Fig. 2.11 gives, for comparison, the maximum productivity of an

equivalent continuous process. Since a continuous process has no dead time,
the slope of the tangent and therefore rj max is greater than in a discontinuous
process (see also Sect. 6.1).
Figure 2.11 illustrates another principle that is an issue in the question of
economic optimization. At point 5, productivity is zero. This point may be of
interest when very expensive substrates are being used. In these cases the
process may be run to complete substrate utilization.
The conversion (i is defined (V = constant)
(. = CI, 0 - cl,t
.1
(2.45a)
Ci . O
and can also be given as a relative quantity
(=_(_i (2.45b)
'i,max
The yield, 1';fj, can be determined by Equ. 2.46 (V = constant); it compares the
total amount of a product j yielded from an amount of material i consumed
_
Yifj- C-J,t - cJ, 0
(2.46a)
ci,o
(cf. yield coefficient, Equ. 2.13) with
1';lj
1';fj,rcl = -y-- (2.46b)
ifj,max
The output of a reactor, with the dimensions [tons/day], can be calculated

from the relation
Output = (i' n i or (2.47)
where ni is the mass flux of component i [t d- 1 ]. In the case (i --> 1 or 1'; --> 1,
the output will be equal to ni .
Point 4 in Fig. 2.11 represents the point where minimum costs (Cmin ) are
reached. The accumulated cost at any time can, in general, be formulated
according to Equ. 2.48 (Richards, 1968a, b)
Cs C
C tot/kg -- W +-
W
R
(2.48)
Costs
C
FIGURE 2.12. Typical change in costs for a discontinuous process over time (Ctot = total
costs, CR = running costs independent of batch size, CB = variable batch costs, which
depend on batch size). t 4 , minimum total costs (see Fig. 2.11). (Adapted from Richards,
1968a,b).
Profit max
'---
/
'"
--
-- --
/
--
--
/ --
--
-t
FIGURE 2.13. Graphic method for evaluating the maximum profit at a given selling
price and starting point, A. (Fixed cost = CB + Ce for deadtime to) (Adapted from
Geyson and Gray, 1972. With permission of John Wiley & Sons, Inc.).
The total costs, Ctot/kg, consist of the sum of the batch costs C B (sterilization,
materials, and costs independent of the amount produced) and the operating
or running costs, CR (for the power used for stirring and aeration). A typical
time course for the accumulation of costs in a discontinuous fermentation
2.3 Systematic, Empirical Process Development With Mathematical Models 41
process is shown in Fig. 2.12. From this, the time when costs are minimal (t 4 )
may be calculated.
The last remaining operating point, point 3 of Fig. 2.11, is that of maximum
profitability. By experience, this point lies between the point of maximum
productivity for a discontinuous process and the point representing minimum
costs.
Profit is calculated according to Equ. 2.49 (Geyson and Gray, 1972)
Profit [$/t] = W (price - Cp,e) - CR' t - (CB + Ce) (2.49)
Cp,c is the extraction cost for product isolation and Ce is the extraction running
cost that is independent of the amount produced. The maximum profit can
be obtained graphically from Equ. 2.49. In Fig. 2.13, the maximum profit can
be obtained from the slope of the tangent to the curve drawn from starting
point A. The use of economic analysis of processing costs is a powerful tool
in establishing priorities for process improvements (e.g., Swartz, 1979; Trilli,
1977).
2.3 Systematic, Empirical Process Development With

Mathematical Models
2.3.1 AN INTEGRATING STRATEGy-A BASIS FOR
BIOTECHNOLOGICAL METHODOLOGY
Research work is often governed in such a way that biology and physics are
elaborated independently of each other. As a consequence, results from the
basic sciences are difficult to apply on an engineering scale for process design.
An integrating strategy has been proposed to bridge this gap (Moser, 1982).
It represents a systematic approach to empirical bioprocess analysis and
design, in which the kinetics of biological reactions play the central role.
The working concept of the macroscopic principle, recently adapted from
chemical engineering (see Roels, 1980; Roe1s and Kossen, 1978), operates with
macroscopically observable variables that are thought to be closely related to
significant phenomena of the process (cf. Table 2.1).
While simplifications are needed, one essential feature of bioprocessing
cannot be neglected: the interaction between microbial metabolism and phy-
sical transport in bioreactors. On the basis of macroscopic process variables,
a systematic analysis of experimental data must include these interactions. As
shown in Fig. 2.14 the fundamental equation of this approach is the law
of the conservation of mass (cf. Equ. 2.3). Macroconversion, re [[, must be
elucidated to elaborate the kind and degree of interactions between meta-
bolism (kinetic rJ and reactor (physical transport, nJ Thus, it is necessary to
quantify bioreactors and their physical transport with the aid of biological
systems having known kinetics ("biological test systems," Fiechter, 1978). It is
also necessary to quantify biokinetics using bioreactors with known physical
transport properties ("perfect bioreactor," Moser, 1978a). Characterization of
EXPERIMENTS
1 to quant i fy
fit
PROCESSKI NET! CS
~j=\-V'ni rj I
PROCESS DESIGN
FIGURE 2.14. Strategies in bioprocess design based on the macroscopic principle and
using the formal kinetic concept as part of an integrating strategy. The strategy
incorporates the spatial change of mass flux through area V'n; (kg/m 2 . h) and the rate
of consumption or formation rj (kg/m 3 . h). rds, rate-determining step; qss, quasi-
steady-state. (From Moser, 1983b.)
biological test systems is mainly gained by quantification: concentration/time

and specific rate/time curves (Dechema, 1982). In a strict sense, these curves
should be elaborated at the level of kinetic modeling.
The characterization of bioreactors is not limited to the standardization of
stirred tank type reactors (Dechema, 1982): It also includes ideal reactors
as model types in a wider framework. The problem of defining a "perfect
bioreactor," including tests of pseudohomogeneity (Moser, 1983a) will be
discussed later.
The interactions of physical transport phenomena and biokinetics, for-
mulated in a sum of process kinetic equations, is qualitatively represented in
Fig. 2.15. The bioreactor behavior on macroscopic level, represented by the
broken line in the figure as a symbol for the area of the balance considered,
can be characterized by calculating derived quantities such as the specific rates
of consumption qs,; growth 11, and production qP,j and formulating the inter-
actions of the kinetic data with the transport of mass, energy, and momentum.
Some other physical properties of the extracellular environment may be
incorporated very usefully as "missing links": viscosity v, shear rate y, density
p, gas hold-up eo, apparent morphology t1* (or M F ), morphology index <5*,
and so on. This approach has been successfully used for the scale-up of, for
-
BALANCE LINE (REACTOR)
.---~----------
.......
I .....--I--p,x
I '---....,..---'
I
I
I
BIOKINETICS
y ..
,...,q~!qS,qc
qHv;qint
MEDIUM
PHYSICAL PROPERTIES
~--"""'-Si
T,pH,rH. p ,EG:E,~*,jt,MF .. ,
P/V,OTR,tc,CTD.RTD. HvTR ...
---- ....... ---- --- -- _...J
FIGURE 2.15. Schematic representation of the concept of interactions between physical

transport phenomena (oxygen transfer rate, OTR; heat transfer rate, Hv TR; power
consumption, PjV; circulation time, t c , and distribution, CTD; residence time distribu-
tion, RTD) and biokinetics (specific rates of growth /1; of consumption of substrates
qs, resp. consumption of oxygen, qo; of production of heat, qH.' of CO 2 qo of product
qp; and of all internal compartments qint). The physical properties of the medium
(temperature T, pH value, redox potential rH, density p, gas hold-up eG , mean energy
dissipation e, shear rate y, resp. "specific viscosity" 1]* = 1]' X-I as a quantitative
estimation of "engineering morphology," and MF = morphology factor) represent the
missing link for modeling the interaction concept. (Adapted from Reuss et aI., 1980.)
1. EXPLORATORY RESEARCH
2. PROCESS KINETIC ANALYSIS
3. PROCESS DES IGN
I
I COMf1ERCIAL REACTOR I - - - - - 1__
FIGURE 2.16. Strategy for engineering analysis of process kinetics according to a

systematic empirical approach. Bioreactors of different scales are used to obtain
process engineering data at different stages (1-3) of development (From Moser, 1978a.)
example, penicillin production (Reuss et aI., 1980) as will be described in Sect.

6.7.
The interactions between physics and biology govern large-scale microbial
production and also heavily influence reaction rates on the laboratory scale.
This fact is to be considered in the case of process kinetics analysis. All
methods used for quantification in bioprocessing (cf. Chapters 3 and 4) should
be critically considered, and standards would be desirable (cf. EFB, Dechema
1984) for development of a general methodology for biotechnology.
In contrast to Fig. 2.2, which demonstrates a process development protocol
without modeling, Fig. 2.16 demonstrates the steps involved in an empirical,
systematic process development that does operate with mathematical models.
Three main areas are associated with the flow of further work: bioreactors,
kinetics, and conversion. This pathway will also be followed in this book
(Chapters 3-6). (Chapter 4 is a detailed look at the problems associated with
the coupling of kinetic and transport phenomena in the formulation of an
analysis of process kinetics.) The goal in each of the three main areas is the
establishment of a mathematical model. In correspondence with the main
objective of this book (see Sect. 1.2), the hallmark of this procedure lies in the
careful formulation of the biokinetics. Basic research is primarily interested
in clarifying mechanisms and in discovering why a process n,ms in a particular

way. Engineering research is primarily interested in questions of know-how-
how a process runs and how it can be influenced so that it runs optimally.
Whether the measured values reflect the micro kinetics depends not on the
scale (laboratory or pilot) but rather on the technique and on whether trans-
port phenomena are at work.
The term "micro kinetics" is understood to mean the kinetics of a reaction
that are not masked by transport phenomena and to refer to a series of
reaction steps. For the investigation of intermediary metabolism, idealized
conditions are chosen that often do not correspond to the real conditions of
engineering processes. This fact makes it difficult to transfer microkinetic data
to technical processes. For the purposes of technologically oriented research
and the development of a process to technical ripeness, it is often sufficient to
know quantitatively how a process runs without necessarily knowing why.
(Macrokinetics, however, must be avoided, as they are scale dependent).
Mathematical formulations are needed that reproduce the kinetics adequately
for the purpose but are as simple and have as few parameters as possible.
Today, even when electronic computers greatly reduce the labor of computa-
tion, the criterion of simplicity remains important due to the problem of
experimental verification. The iterative nature of the process of building an
adequate model is an important point that will be considered in greater detail
in Sect. 2.4.
The successive working stages of analysis and synthesis characterizing the
flow of work in a systematic process development are shown schematically in
Fig. 2.16. A distinction is made among three levels:
- Fundamental research in small laboratory reactors (Erlenmeyer flasks)
- Process kinetic analysis in laboratory-scale reactors (V = 10-501)
- Process design in pilot plant reactors
The process kinetic analysis is carried out in two different types of reactors.
The so-called "perfect bioreactor" is used for obtaining "true" kinetic data.
This laboratory reactor must therefore meet certain requirements with regard
to all possible transport phenomena (see Sect. 4.2). A so-called "bioreactor
model" is a scaled-down bioreactor with some geometrical similarity to the
production unit. Here, significant transport phenomena (cf. Chap. 3) can be
studied and quantified with a physical system in a first step.
Pilot plant bioreactors serve one purpose supplementary to those men-
tioned before (see Sect. 2.1.4): test of the process model (process kinetics),
including separate tests of kinetics and physical transport phenomena.
Finally, there is a precautionary rule: Kinetics should always be measured
in two different reactors to identify any hidden variables or parameters. In
addition to experimenting in a perfect bioreactor, one should also evaluate a
trial in the pilot plant. This testing has another purpose as well, which might
be called "kinetic similarity" (Moser, 1978b) (as exists, e.g., between DCSTRs
and CPFRs). Real biological processes are so complex that altered flow
conditions cause changes in the concentrations of materials and correspond-

ing changes in the metabolism and constitution of the mixed population of
cells (biocoenosis). The advantage of kinetic similarity is that it guards against
the eventuality of unexpected factors not considered in the model. Perhaps
with further progress in building reliable models this step may no longer be
necessary.
2.3.2 WORKING PRINCIPLES OF BIOPROCESS TECHNOLOGY
Based on the strategy of systematic process development, an engineering

approach to bioprocessing includes several stages, as illustrated in the scheme
of Fig. 2.14. As already stated in Fig. 1.4, the concepts of a rate-determining
step (rds) and of the quasi-steady-state (qss) are very useful (see Sect. 2.4.1).
Some basic principles play an important role: to simplify, to quantify, to
separate, to model, and to recombine the separated phenomena to form a total
process model.
2.3.2.1 Principle of Simplification: The "Formal Macroapproach"

The engineer responsible for planning must always strike a balance between
involved complications and practicality; in addition every model will be
characterized by the number of variables and the measuring techniques used
for them. The predictive power of the model will thus be limited.
The following statement will serve as a general guideline for the first
working principle: Simplification includes compressing the complex structures
of a process into those few schemes that may be regarded as significant, that
is, that are reflected in key variables. In this way both the experimental and
theoretical research may remain limited to just that degree of precision neces-
sary for the purpose at hand.
Due to the complexity of both aspects-the biological and the physical-
simplifications must be made without loss of essential information. This
approach, well known in chemical engineering, is a consequence of the macro-
scopic principle (Glansdorff and Prigogine, 1974), which has recently been
applied to bioprocessing (Kafarow et aI., 1979; Moser, 1978a, 1981; Roels,
1980a, b; Roels and Kossen, 1978; Votruba, 1982). In other words: "Everything
should be made as simple as possible, but not simpler," as Einstein said.
According to Fig. 2.3, microbial cells suspended in the liquid phase of the
bioreactor are treated as black boxes; this does not mean that they are
neglected. Their macroscopic behavior, manifested in the changes of the
concentrations in the liquid phase, is taken into consideration. As a con-
sequence of this approach, the bioreactor itself is not treated as a black box.
The core of the formal macroapproach is that formal analogies are used (see
Sect. 2.4.3). The utility of the principle of simplification can also be demon-
strated in the case of modeling the dynamics of bioprocesses (see Sects. 3.5.3
and 5.7).
2.3.2.2 Principle of Quantification

Quantitative data collection is widely accepted as the basis of every modern,
effective technology. Section 2.2 laid out the concepts for data collection for
bioprocesses. Analytical methods for measuring process variables are often
faulty, that is, they may be falsified by interactions between physics and
biology. The question of the extent to which the measured variable is a useful,
representative variable is generally of far greater importance, thus the analy-
tical methods as used, and this is especially so in evaluating the value of a
mathematical model. Examples where this is an issue include (a) the dry cell
mass used as a substitute for the catalytically active cell mass and (b) the
biological oxygen demand as a measure of the degree of pollution in waste
water treatment.
2.3.2.3 Principle of Separation

Separation deals with the accurate design of experiments for obtaining data
on biological and physical phenomena in circumstances where the rates of the
biological and physical processes are independent of one another. Macro-
conversion (macrokinetics) is an ill-defined basis for scale-up, and efforts must
be made to avoid the appearance of such data. This is done with the aid of
the test of pseudo homogeneity (see Sect. 4.3) and of regime analysis in general
(see Sect. 4.2).
In connection with the problem of pseudohomogeneity, it is clear that a
reaction occurring inside a solid phase is not directly measurable. The internal
flux of a metabolic reaction can thus only be checked by measuring the
external fluxes in the liquid medium by means of computer simulation. Bar-
ford and Hall (1979) found that an external overall flux does not reflect, even
in an approximate manner, the internal fluxes. Moreover, in vitro examination
of an isolated section of metabolism is inadequate for quantification of the
coordinated and integrated biochemical control of an intact living system due
to in vivo interactions between different parts of the metabolism.
Another application of the principle of separation is the case of truly
heterogeneous processes (see Sects. 4.5, 6.6, and 6.7). Here, cases of external
and internal transport limitation of a bioprocess must be distinguished from
the case where the physical transport may be enhanced by the biological
reaction. A simplified treatment in all these cases is possible through introduc-
tion of the concept of efficiency IJn defined by
(2.50)
The efficiency factor of the reaction IJr can be interpreted as
(2.51 )
being a function of the interaction between rate constant of the reaction kr

and transport kTr (see Sect. 4.5). With the concept of efficiency, the real
heterogeneous situation can sometimes be handled as a pseudo homogeneous
case.
The principle of separation constitutes an important prerequisite for the
sensible application of mathematical models, and vice versa.
2.3.2.4 Principle of Mathematical Modeling

Mathematical models are mathematical formulations able to represent the
behavior of a process on a simplified level and adequately for a given purpose.
Such models serve to generalize singular results and provide a basis for the
deduction of further properties of the system. Process and system engineering
use mathematical models and a high degree of theoretical background; this is
often admired, but alsD often rejected by biologists. However, Louis Pasteur's
admonition should be kept in mind: "Without theory, practice is but routine
born of habit." This statement is related to deductive research methodology,
which is accepted as the basis for scientific work (Popper, 1934, 1972). The
steps of deduction in bioprocess analysis and design are illustrated in Fig. 2.17.
As theories are always fallible, the only way to circumvent failure is to compare
hypotheses with experiments. To be sure that no undetected interaction is
disturbing the measured kinetics, a supplementary run should be made in a
pilot plant reactor. Here transport phenomena are normally quite different
from those of the bench-scale reactor, so that interactions are easier to detect
and modifications to the model can be made, such as the addition of new
model functions j; other process variables Xi' or other model parameters k.
All we can do is to apply the principle of analogy, describing the behavior of
the system under certain conditions ("if, then"). These "black-box" models,
however, are not only descriptive but also predictive when the deductive
method is applied.
Even if some mechanistic background allows a deeper interpretation of the
parameters so that they have more biological significance, with such an
approach "gray boxes" still remain; the approach is still, more or less, an
optical illusion (Roels and Kossen, 1978), since the final subsystems are always
black boxes. In conclusion, the level of modeling to be considered as adequate
depends on the principle of analogy, in which the model parameters are
adapted to experiments ("adaptational parameters," according to Hofmann,
1975).
To summarize the principal aims of mathematical models:
1. Models are built to allow prediction of the conversion or productivity of
any system.
2. Models are built to examine the nature and behavior of a plant's operation
under a variety of operating conditions and to examine the regions where
the model is valid, including how far from this region extrapolation is
permitted.
2.4 Mathematical Modeling in Bioprocessing 49
3. Models allow generalization to other situations within the boundaries of

their validities.
4. Models are mathematical formulas that can be manipulated to optimize
processes.
5. Computer simulations are types of mathematical models.
6. Models are used to identify unknown or previously disregarded process
variables and parameters that may be significant variables.
7. Models serve as controls in examining whether an effective separation of
biological and physical phenomena has been achieved.
8. As an indirect result, models may also help in clarifying reaction mechanisms.
2.4 Mathematical Modeling in Bioprocessing

Because of the central meaning of working with mathematical models, their
general nature, objectives, and construction will be discussed here.
2.4.1 GENERAL REMARKS

In the literature there are several ways of classifying different types of models
(cf. Roels and Kossen, 1978, and Sect. 2.4.3). From the viewpoint of process
engineering, in this book distinctions will be made among macro kinetic,
microkinetic, formal kinetic, and process kinetic models (see Fig. 2.14). The
formal models that will be discussed individually in Chap. 5 are sometimes
referred to in the literature as "unsegregated" or "unstructured models" that
have a "descriptive" or "predictive" nature.
Classification by the concepts of "segregation" and "structuring" (Tsuchiya,
Fredrickson, and Aris, 1966) is predominant in the area of microkinetic
models, but it is also useful for making broad distinctions among formal
kinetic models. Unsegregated models are those that consider the biomass of
each cell as identical in terms of physiology, morphology, and genetics. The
number of cells involved in most fermentation processes of technological
importance is very large. Within the limits of the precision desired for an
engineering process, the macroscopic behavior can be treated as constant,
representing an average value throughout the whole population. On the other
hand, unstructured models neglect the fact that the cell mass is not uniform
but rather consists of many different compounds (DNA, RNA, proteins, etc.)
that, kinetically, behave quite differently. The problem of structuring mathe-
matical models will doubtless become more important, especially in the case
of nonstationary processes such as batch and fed batch cultures (cf. Sect.
5.7.2.2).
It cannot be expected that any kinetic model will be directly applicable to
a real process situation. Mathematical modeling must start with the simplest
type, but it must be reiterated, modified, and extended until eventually it leads
to an adequate process kinetic model (see Fig. 2.17). In modeling, great value
r-------- EXPERIr1ENTS
I
I
I
I
I
I
I
I
I _ _ _ __ feff
I ANALYSIS
I '-------,r-r' SINGULAR CASE DATA
I
II BIOLOGY:
I KINETICS
I fj
HYPOTHESES FOR
I STO I CH I OMETRY
I v. Y.. SUBMODELS
I Ij IJ
I
I
I DEDUCT I VE 11ETHOD
I
I I NDUCT I VE ~'ETHOD
- - - - - - - - - - - PROBLEM
I\/n.< rI'
I I
f(X,k)
~
SIGNIFICANT
PROCESS VARIABLES ~i
PROCESS f10DEL
GENERAL VAll 0
FIGURE 2.17. Flowchart of adaptive modeling in the integrating strategy, applying

deductive research methodology in which mathematical models (for physical trans-
ports l1i' biokinetics ri , and stoichiometry on the macroscopic level 1';u or the micro-
scopic level ViU) serve as working hypotheses to be compared consequently with
carefully designed and evaluated experiments. Special attention concerns the interac-
tions between 11;, ri , and 1';u' as indicated. (From Moser, 1981).
should be placed on establishing the boundaries within which the mathe-

matical functions are valid and the actual values to be used for the parameters
in concrete instances of application.
In connection with this discussion it might be said that a rather simple
method of structuring, such as, for example, distinguishing between the quan-
tity of protein and the quantity of nucleic acids present, can lead to a great
number of otherwise inaccessible parameters (Reuss, 1977). This can make an
important contribution to the understanding of intracellular events. For
engineering calculations, however, the primary claim of simplicity must also
be satisfied. Furthermore, within the scope of the analytical methods pre-
sently available, distinctions among models are often simply not possible
(Boyle and Berthouex, 1974).
The important characteristics of models to be considered for engineering
purposes are
1. The model presents a simplified picture of the complex structure of a process

in terms of a few key variables.
2. The model is developed to serve a particular need (generalizations can be
achieved by "putting the model in jeopardy").
3. The model forms a bridge across the gap between microscopic and macro-
scopic phenomena.
4. The model has been "verified"-that is, it has been compared with experi-
mental results and adapted to them.
5. The model is valid only within certain boundaries, which correspond to
the models purposes. No model can completely replace experimentation.
Models can, however, contribute to saving some of the time and money
necessary for experiments. "Parameter sensitivity analysis" is obligatory.
2.4.2 MODEL BUILDING
The mathematical formalism of Equ. 2.2a presents one case of a method for
reproducing kinetics. The general form of a mathematical model is
(2.52)
In accord with both theory and experience, the total function can be separated
into a function of temperature and a function of the concentration variables:
r = k(T) f(c x ) (2.53)
This very general equation is the starting point for developing an analysis of
process kinetics (see Chap. 4).
2.4.2.1 General Procedure

Model building is concerned with the establishment of equations of this type.
Important considerations in doing this are
- The choice of the significant process variables, Xi.
- The choice of an appropriate mathematical function for the model, f, and
of the appropriate fitting parameters, k.
- Comparison of the first model with experimental results and the expansion
or modification of the model functions.
The stages of this procedure are shown in Fig. 2.1S. The initial motivation
is always the existing concept of the purpose of the model. On the basis of
information from the literature and/or exploratory experiments (and with, of
course, the element of intuition), the problems can be recognized and a
hypothesis formulated in mathematical terms. A model is selected by first
selecting the significant process variables and then preparing a qualitative
structure diagram in the form of a simplified process flowchart. The next phase
can be described as one of testing or verification. It consists of using estimated
test data to obtain a mathematical solution to the system of equations, to
EXPERIM.
DESIGN
rN-E-W-S-r-G-N-rF-I-CA-N-T-v-A-R-rA-B-L-ES--!~ Choose 5 I GN I F I CANT PROCESSVAR IABL ES Xi - - - - t
Write QUALITATIVE PROCESS-SCHEME
NEW Ba 1ance
FUNCTIONS
Quantify KINETIC MODEL EQUATlDNS
no
yes
Compare
Modif MODEL E UATIONS no no

accuracy.
no
MATHEMATICAL MODEL of
FORMALKINETICS
FIGURE 2.18. Procedure for constructing a formal kinetic mathematical model of a

bioprocess (Moser, 1977c.) Explanation see text.
determine whether the model appears, in principle, to reproduce the course

of the process.
Identification with the model parameters is accomplished in most cases with
use of graphically estimated linearizations or of nonlinear regression analysis.
After model identification, parameterization may be undertaken. In the begin-
ning the control of parameterization should not be delegated to the computer
operating blindly with prescribed criteria. Deviations can suggest a previously
unknown influence that should be incorporated into the model.
A "sensitivity-analysis" has to show finally how the model reacts to chang-
ing values of parameters. An efficient biokinetic experimental design has been
demonstrated by Johnson and Berthouex (1975a), and multiresponse data are
used for parameter estimation by the same authors (1975b).
Three factors are possible sources of error in constructing a model (Moser,
1978b):
1. Process variables not adequately considered, for example, O 2 , CO 2 , visco-

sity, T temperature pH, inhibitory metabolites.
2. Choice of inadequate submodels of the kinetic processes, for example,
unknown influences from endogenous metabolism, inhibitors, the use of
homogeneous rather than heterogeneous models, multisubstrate limita-
tions, mixed population interactions.
3. Experimental requirements or assumptions concerning transport not ful-
filled, for example, the mixing behavior of fluids and gases, deviation from
the ideal residence time distribution, critical dependence on particle size,
the thickness oflayers of cells, the response lag of the P02 electrode.
Searching for the source of error is mainly an intuitive process, but one that
is aided by close observation during the course of experiments and close
adherence to protocols. Model building has to be repeated several times.
The iterative nature of the "adaptive model building" process is of special
importance not just for effective parameterization (Johnson and Berthouex,
1975a, b) but also for the whole process of conceptualization and working with
mathematical models of kinetics.
2.4.2.2 Integration of Balance Methods and Kinetics in the Construction

of Unstructured Models
To construct a simple unstructured model for bioprocesses, at least one of
the reactions taking place in the culture must be specified in kinetic terms.
Generally a complete set of constitutive equations for each of the N chemical
reactions taking place in the culture can be written in the form of a sum or a
matrix (Roels, 1980a; Schubert and Hofmann, 1975). The net conversion rate
of each of the components present follows with the aid of rj = vj r. For a
system in steady state, the net production rate is equal to minus the flow into
the system, as is clear from Equ. 2.10. Furthermore, the elemental balance
principle (according Equ. 2.11) specifies k relationships between the flows Fj
DECIDE N SIGNIFICANT
PROCESS VARIABLES PROCESS KNOW HOW
14--CREATIVE THINKING
EQ. 29,211
LITERATURE
CREATIVE THINKING
EQ.2-3A
FIGURE 2.19. Flowchart of the construction of unstructured biokinetic models as a

consequence of joint activity between kinetic work and stoichiometry. (Adapted from
Roe\s, 1980a. With permission of John Wiley & Sons, Inc.).
and hence only (N - k) net conversion rates can be chosen independently.

The number of independent kinetic equations to be postulated cannot be
chosen at will-it is completely specified by the number of elemental balances
(k) and the number of components in the system (N). The procedure for the
construction of a simple unstructured process model is outlined in individual
steps in Fig. 2.19.
From the foregoing it is clear that the theory of elemental balance is an
invaluable tool in the description of the systems commonly encountered in
bioengineering. It is as fundamental as stoichiometry in chemical reaction
engineering systems. The theory seems to be well developed, and the field is
open to the development of specific applications. A significant problem is,
however, associated with the application of the theory. It applies to instances
in which more flows are measured than are minimally needed to calculate the
remaining ones. In this case, a statistical procedure can be applied to obtain
a more optimal estimate of all measured and unknown flows.
2.4.3 DIFFERENT LEVELS AND TYPES OF KINETIC MODELS
Some remarks should be made here to clarify discrepancies observed in

the kinetics literature. Classification of mathematical models of microbial
processes normally involves contrasting pairs, for example: continuous time-

discrete time, deterministic-stochastic, unsegregated-segregated, unstructured-
structured, lumped parameter-distributed parameter (Harder and Roels,
1981; Kossen, 1979; Reuss, 1977; Roels and Kossen, 1978), to USe terminology-
from Tsuchiya et ai. (1966). The kinetics of biological systems may be
expressed at four different system levels: (a) molecular or enzyme, (b) macro-
molecular or cellular component, (c) cellular, and (d) population (Humphrey,
1978). Each level of expression has a unique characteristic that leads to a rather
specific kinetic treatment, similar to the situation with chemical reactions. The
similarity to chemical processing should be emphasized even more when the
high complexity of bioprocessing at the engineering scale is considered. In
complex systems, the mutual relations among phenomena are clarified by
means of systems analysis (Kafarow, 1971). Knowledge of Macrokinetics-
as named by van Krevelen in his introduction to the first symposium on
chemical reaction engineering (1957)-is a sine qua non for studying conver-
sions and developing reactor operations. It is only by understanding the
relation and the interaction between micro kinetics and macro kinetics that the
real phenomena occurring in a chemical or biochemical/biological reactor can
be understood.
On the basis of empirical data from reactor operations, and with use of
systems analysis, a general research methodology can be derived (Fig. 2.20).
The process need not be examined in its infinite complexity but can be
hierarchically subdivided into five different levels. The fourth level of bio-
reactors is the most essential, according to the macroscopic principle. But all
other levels contribute to the development of bioplants according to the
special aims of the investigation. Generally, model building can proceed in
various ways:
1. Derivation of an equation from the microkinetics of the reaction mechanism
2. Simulation of experimental curves by means of purely numerical methods
3. The establishment of quantitative formulas with the help of formal analogies
Each of these three ways of constructing models will be discussed separately.
2.4.3.1 Mechanistic Models of Microkinetics

In principle it is possible to investigate the biochemical mechanisms of the
metabolic pathways (enzyme induction and metabolic repression). In many
cases a good beginning is described in the literature (Moreira et aI., 1979).
From this start, a mathematical function usually can be formulated that
possesses such a high degree offlexibility that it can be brought into agreement
with the experimental facts. In most cases, however, good ideas about the
biochemical regulation of metabolism are all too few, and exact values for the
parameters of the model are not available. One must then deal with estimates,
often estimates taken from the literature.
The complex efforts of modeling also serve another end, namely that of
process control and special process optimization (Calam, Ellis, and McCann,
1971). For this, naturally, special methods of analysis must be available. In
UNIT OPERATION CONCEPT

5
( PREPARATJ ON, REACTOR, RECOVERY)
BlOPLANT
MODEL ECONOMICS
OPTlr1IZATJON
4 r1ACRO- KINETI CS
BlOREACTOR
t10DEL
(BIOPROCESS KINETICS) IN DIFFERENT BIOREACTOROPERATJON
YIELD CONCEPT
3
POPULATION PURE / MIXED POPULATIONS
~lODEL
MORPHOLOGY
( FLOCS,MYCELlA,PELLETS)
2
MICRO -
SINGLE CELL
f10DEL
COMPLEX MET ABOLI C NETWORK
CELLULAR COMPONENTS
COUPLI NG B I OSYNTHES IS/ENERGY
CATABOL / ANABOL REACTIONS

fl0LECULAR LEVEL
ENZYME ACTIVITY CONTROL
f10DEL
GENETICS
BASIC STOICHIOMETRY
FIGURE 2.20. Overview of the five levels of process research and mathematical modeling
of bioprocessing and bioreactor operation.
the future, the microkinetic approach will no doubt contribute a great deal to
the formulation of process kinetics (see App. II, especially the use of rds
and qss).
The derivation of kinetic equations from postulated mechanisms can, for
instance, show the analogy between enzyme kinetics and chemical kinetics for
heterogeneous catalysis. Poison-free enzyme kinetics follow the form of a rate
equation (Michaelis-Menten type)
s
rs = rs.maxK (2.54)
m+ S
while for heterogeneous catalysis, the Langmuir type is useful (with c = s)

KLc
, = 'max 1 + KLc (2.55)
Evidently, both approaches are identical when
(2.56)
(i.e., the Michaelis- Menten constant is inversely proportional to the adsorp-

tion equilibrium constant).
Engineers are mainly interested in manipulating a population of celis,
and fundamental biology can contribute significantly to deriving population
models from single-cell models, this has been pointed out by Shuler and
Domach (1982) at a symposium on kinetics and thermodynamics in biological
systems (Blanch et al., 1982). Single-cell models can be formulated that mimic
very closely the responses of living cells. Such models are convenient tools for
testing the plausibility of basic biochemical mechanisms. They may be parti-
cularly attractive for testing the potential in vivo compatibility of mechanisms
postulated on the basis of in vitro enzymology. Fig. 2.21 shows an idealized
sketch of a single-cell model (the Cornell single-cell model, Shuler and
Domach, 1982), representing a modification of a previous simple mathe-
matical model for the growth of an individual bacterium (Ho and Shuler,
1977). The simple mathematical model could predict the growth pattern for
a cell of a given shape (filamentous, bacillus, or spherical), and it initially
included four components (NH;, glucose, precursors, and macromolecules).
This number was later enlarged to 14 components (Shuler et aI., 1979). The
Cornell single-cell model makes reasonable predictions about the dependence
of cell size, cell shape, cell composition, growth rate, and timing of cellular
events on external concentrations of glucose.
More complex models with a larger number of parameters are able to
describe real processes better than simple models. For practical engineering,
however, models should be as simple as possible but no simpler (i.e., with a
minimum number of parameters).
The field of kinetic modeling of nonstationary bioprocess situations with
the aid of structured and mechanistic models will be discussed in more detail
in Chap. 5.
2.4.3.2 Numerical Fitting

Equ. 2.57 shows a purely mathematical function that is well suited for repro-
ducing the time-dependent growth curve of a discontinuous process. It con-
tains both an exponential term and a polynomial, which makes it very flexible
(Edwards and Wilke, 1968).
Type 1: rj* = f(t), Example
K
,.* = ----------0---- (2.S7a)
1 1 + exp(iX
o iXIt Z+ + iXzt + ... )
-1 -2
Vl
00
IV
~ -l
::r
(1)
~ \~M5
GIS '"C
::!.
::s
(")
ii'
'
'\~ 13 ir
o
1 - - - 12
1 - - - ..... \
\,
\
I
t
'., 00
.....
" .... " ....,
t:t'
o
PG .. '0
. --.....~II
.... / ' /
a
/' / ...... ....
I - -; --..""
M1 ,./
---- - - - M2 - "
~
CIl
~M ' M2RTM , M2RTI

~
::s
o
0-
OQ
'<
M4
FIGURE 2.21. An idealized sketch of the model for Escherichia coli ing 85% r - RNA throughout); M 2M , messenger RNA; M 3 , DNA; M 4 ,
growing in a glucose-ammonium salts medium with glucose or nonprotein part of cell envelope (assuming 16.7% peptidoglycan,
ammonia as the limiting nutrient. At the time shown, the cell has just 47.6% lipid, and 35.7% polysaccharide); Ms, glycogen; PG, guanosine-
completed a round of DNA replication and initiated cross-wall fer- tetra phosphate; E I , enzymes in the conversion of P 2 to P 3 ; E 2, E3 ,
mentation and a new round of DNA replication. Solid lines indicate molecules involved in directing cross-wall formation and cell envelope
the flow of material, while dashed lines indicate flow of information. synthesis (the approach used in the prototype model was used here but
SI' ammonium ion; S2, glucose (and associated compounds in the cell); more recent experimental support is available); G, glutamine; E4
Pi' waste products (C02, H 20, and acetate) formed from energy glutamine synthetase. An * indicates that the material is present
metabolism during aerobic growth; II, amino acids; 12, ribonucleo- in the external environment. (From Shuler and Domach, 1982. Re-
tides; 13 , desoxyribonucleotides; 14 , cell envelope precursors; M I, printed with permission from Kinetics and Thermodynamics in Bio-
protein (both cytoplasmic and envelope); M 2RTI , immature "stable" logical Systems, ACS-Winter Symposium. 1982 American Chemical
RNA: M w r u mature "stable" RNA (r - RNA and t - RNA; ass um- Society).
with rt = Ji, K = X max , et o= xmax/XO' and et 1 = Jimax. The disadvantage of the

usually anonymous functions is that the parameters are mainly devoid of any
biological meaning. Other simple functions are also used to describe the
time-dependent curve, for example
(2.S7b)
The same type can be used to describe the curves for the concentration
dependence r = fCc) and/or r = f(t). In principle, the same function as that in
Equ. 2.58, or even a simple polynomial, may be used.
Type 2: rt = fCc): Example
(2.58)
The limitations that apply to this type of analysis do not necessarily diminish
its value. Applications will remain limited to cases of very difficult to analyze
processes and to the initial phases of process development (e.g., Grm, Mele,
and Kremser, 1980). Sooner or later, when economic incentive is adequate,
mathematical models with parameters interpretable in clear biological terms
will be preferred. The well-known kinetic approach of Kono (see Chap. 5.3)
is also a numerical fitting procedure of great engineering value.
2.4.3.3 Formal Kinetic Approach

Here, thinking in terms of analogies arises as a fundamental principle very
useful for process development but also having dangers. A formal kinetic
model contains analogies to known precedents that are represented by mathe-
matical formulas, formula that allow a simple (and therefore usable) reproduc-
tion of the course of a process, at least those aspects regarded as significant.
A formal kinetic approach binds the user to a somewhat distorted and
mechanistically causal interpretation. This is not expressly stated in the
approach and may be of no importance with regard to the primary purpose
at hand. However, the implication is that the microkinetic reality is somehow
mirrored in the formal kinetic analysis (cf., e.g., Fig. 5.47). Despite the progress
of fundamental sciences in the field of biology, especially in the development
of models with a mechanistic background, the useful mechanistic interpreta-
tion of a bioprocess with even normal complexity is still a demanding task.
In this situation, real experiments should be the starting point of process
kinetic analysis (Moser, 1978a) as in Fig. 2.14.
The special definition of formal kinetics can be summarized as follows
(Moser, 1983b):
1. The experimental data and not postulated mechanisms are the starting
point.
2. The mathematical function to be chosen as the model is not of the purely
numerical form but is taken from analogies. Analogy means that known
mathematical functions, generally from conventional mechanistic work
with a similar phenomenology, are selected (e.g., the Monod equation as
an analogy to enzyme kinetics).
3. A representative quantification of the process is reached by adapting the

values ofthe model parameters to experimental data (so-called adaptational
or fitting parameters e.g. Ks).
4. The meaning of the parameters of such formal kinetic models is different
from that of real mechanistic models. A degree of mechanistic interpretation
can be attained only with supplementary work. This fact is the main
difference between formal kinetics and micro kinetics; the mathematical
structure used in bbth approaches often is the same (e.g. Ks # Km).
This formal kinetic approach is also successfully used in chemical kinetics
(Hofmann, 1975), and it is advantageously used for complex chemical reac-
tions (Schmid and Sapunov, 1982).
A systematic formal kinetic analysis starts with measured concentration-
time curves (e.g., in batch processes, as illustrated in Fig. 2.4 for substrate
concentrations). From these data a reaction scheme can be extracted. At this
point a clear differentiation must be made between reaction scheme and
reaction mechanism. Due to the fictitious character of a mechanism, it may
be dis proven but never proven. A reaction scheme, on the other hand, can be
more or less definitely established and may be extended later only if there is
evidence of additional steps. From the shape ofthe concentration-time curves
several conclusions can be made (Moser, 1983b) concerning the interpretation
of apparent reaction orders n. Linearity can be a sign for transport limitation
or can indicate the presence of a biosorption effect resulting in a reaction order
of zero. Half- and first-order reaction can be interpreted as internal transport
TABLE 2.5. Significant phenomena in bioprocessing accord to

the formal macroapproach, their quantification, and pseudo-
homogeneous modeling.
Macroscopic Observed Derived Model
phenomena quantity quantity parameters
Kinetics
Growth x r x , Jl J1.max' Ks
S consumption s rs, qs qs,max' Kg, Y,qS
O 2 consumption 0 ro' qo qO,max, Ko
Product formation p rp, qp qp.max, YP1X , kp
CO2 formation c rc , qc qc,max
Heat formation hy rH qH. qHv.max, k'X ~ Ea
Maintenance cj kd' m" k f.i.d
Stoichiometry Cj YX1S , YXIP '
Transports
L phase
O2 0 no (OTR) OkLlQ,OkL2Q
CO2 C ne(CTR) C kIIQ
Heat hy T q(HyTR) kH.aH

Micromixing cj t m, te. CTD t m , tc
Macromixing cj 1, RTD S2 (80, N )
S phase dp d erit dp(Deff)
Adapted from Moser, 1981.

limitations (see biofilm kinetics, Chap. 5.5), while a changing order from zero
to one is to be expected for enzymatically controlled bioprocesses. However,
the magnitude of the model parameters, especially of the Ks value, can be
taken as an indicator for different overlapping phenomena (Fiechter, 1982;
Moser, 1981; see Chap. 5.9). Table 2.5 summarizes the measurable process
variables considered significant. The working principles discussed in this
chapter are also apparent in this table. The last column identifies the param-
eters that are incorporated in the case of a mathematical model of the kinetic
and transport p r o c e s s e s . ,
Finally, the field of scientific activities for the set-up of mathematical models
is shown by illustrating the real situation in biopressing (see scheme of Fig.
2.22). The explanation of details given in the legend to this figure is intended
to serve as an introduction to the next chapters.
------........
I
/ 0
I
,I
0
kE ,
0 --I,P
/
j
I
"
---5 : I
\0 ,f
\ 0 I
"" ,I
.-
, ,o-
0
f
\
oG kL

I
~ Tr
,,
'-.
I(;
n
FIGURE 2.22. Schematic representation of the basic experimental situation in bio-

process/bioreactor analyses, where the interactions between physical transports (k}r)
and biokinetic rates (k~) in the liquid phase are thought to be representative for the
process rates in the solid phase of cell mass (k~). At the same time, response lags of
measuring electrodes (kE) have to be taken into account. G, gas phase; L, liquid phase;
S, solid phase or substrate; E, enzyme or electrode; I, intermediary metabolites or
products; P, end product; N, nucleus; R, ribosomes; M, mitochondria; IX, anabolism;
p, catabolism; Fo = gas flow rate; n = agitators rotational speed.
BIBLIOGRAPHY
Abbott, B.J., and Clamen, A. (1973). Biotechnol. Bioeng., 15, 117.

Agrawal, P., et a\. (1982). Chem. Eng. Sci., 37,453.
Aiba, S., Humphrey, A.E., and Millis, N.F. (1976). Biochemical Engineering. New York:
Academic Press.
Atkinson, B., and Mavituna, F. (1983). Biochemical Engineering and Biotechnology. The
Nature Press and McMillan Pub\. Ltd.
Bailey, J.E. (1973). Chern. Eng. Commun., 1, 111.
Bajpaj, R.K., and Reuss, M. (1981). Biotechnol. Bioeng., 23,717.
Barford, J.R., and Hall, R.J. (1979). In Proc. 7th Australian Con! Chem. Eng. p. 21.
Bauchop, T., and Elsden, S.R. (1960). J. Gen. Microbiol., 23,457.
Bazin, M.J. (ed.) (1982). Microbial Population Dynamics. Boca Raton, Fla.: CRC Press.
Bell, G.H. (1972). Proc. Biochem., 7 (April), 21.
Blanch, H.W. et a\. (eds.) (1982) ACS winter symp. "Kinetics and Thermodynamics in
Biological Systems" Boulder, Colorado.
Bogen, H.J. (1976). Buch der Biotechnik. KnaUf Verlag, no. 3478. Munich-Zurich:.
Boyle, W.C, and Berthouex, P.M. (1974). Biotechnol. Bioeng., 16, 1139.
Budde, K., Budde, H., and Riickauf, H. (1981). Stoichiometrie chemischtechnologischer
Prozesse, Berlin: Akademie-Verlag.
Cal am, CT., Ellis, S.H., and McCann, M.J. (1971). J. Appl. Chem. Biotechnol., 21,181.
Cooney, Ch.L. (1979). Proc. Biochem., 14 (May), 31.
Cooney, Ch.L., and Alcevedo, F. (1977). Biotechnol. Bioeng., 19, 1949.
Cooney, Ch.L., Wang, H.Y., and Wang, D.I.C (1977). Biotechnol. Bioeng., 19,55.
Dechema Monograph (1982). Arbeitsmethoden for die Biotechnologie. Biotechnology
working group (chairman H.J. Rehm). Frankfurt: Dechema.
Dechema Monograph (1984). Process Variables in Biotechnology. Bioreactor Perfor-
mance EFB-working party (chairman W. Crueger) Frankfurt: Dechema.
Dostalek, M., Haggstrom, L., Molin, N., and Terui, G. (1972). In Ferm. Technol. Today,
Proc. 4th Intern. Ferm. Symp. (Terui G., ed.) Society of Ferm. Techno\. Japan, 497.
Edwards, H.V., and Wilke, Ch.R. (1968). Biotechnol. Bioenq., 10, 205.
Endo, I., and Inoue, I. (1976). 5th International Fermentation Symposium, Berlin,
Abstract 5.01. Dellweg H., ed., Inst. f. Garungsgewerbe und Biotechnologie.
Erickson, L.E., et a\. (1978a). Biotechnol. Bioeng., 20,1595.
Erickson, L.E., et a\. (1978b). Biotechnol. Bioeng., 20, 1623.
Erickson, L.E., et a\. (1979). Biotechnol. Bioeng., 21, 575.
Eroshin, V.K. :1977). Proc. Biochem., 12 (July/August), 29.
FAST (1980). Sub-programme Bio-society research activities, Commission of the
European Communities, EUR 7105, Brussels (M. Cantley).
Fiechter, A. (1974). In Proc. 4th Intern. Symp. Yeasts, Klaushofer H. and Sleytr U.
(eds.), Univ. of BodenkuItur Vienna, Part II, 17.
Fiechter, A. (1978). In Proceedings of the 1st European Congress on Biotechnology,
Interlaken, Switzerland Dechema Monograph 82,17. Dechema ID.
Fiechter, A. (1982). In Rehm, H.J., and Reed, G. (eds.). Biotechnology-A Comprehen-
sive Treatise, Deerfield Beach, Fla and Basel: Verlag Chemie Weinheim. Vo\. 1,
Chap. 7.
Geurts, Th.G., et a\. (1980). Biotechnol. Bioeng., 22, 2031.
Geyson, H.M., and Gray, P. (1972). Biotechnol. Bioeng., 14,857.
Glansdorff, P., and Prigogine, I. (1974). Thermodynamics of Structure, Stability and
Fluctuations. New York: John Wiley.
Grm, B., Mele, M., and Kremser, M. (1980). Biotechnol. Bioeng., 22, 255.
Guenther, K.R. (1965). Biotechnol. Bioeng., 7,445.
Haggstrom, L. (1977). Appl. Environ. Microbiol., 33, 555.
Harder, A., and Roels, J.A. (1981). Adv. Biochem. Eng., 21, 56.
Harrison, J.S. (1967). Proc. Biochem., 2, 41.
Heijnen, J.J., and Roels, J.A. (1981). Biotechnol. Bioeng., 23, 739.
Heijnen, lJ., et al. (1979). Biotechnol. Bioeng., 21, 2175.
Herbert, D. (1975). In Dean, A.C.R., et al. (eds.). Continous Culture, Vol. 6. London:
SCI, E. Horwood Ltd, Chichester England p. 1.
Ho, S.V., and Shuler, M.L. (1977). J. Theoret. BioI., 68, 415.
Hockenhull, D.J.M. (1971). Progr. Ind. Microb., 9, 133.
Hockenhull, DJ.M. (1975). Appl. Microbiol., 19, 187.
Hofmann, H. (1975). Chimia,29, 159.
Hoover, S.R., and Porges, N. (1952). Sewage Indus. Wastes, 24, 306.
Humphrey, A.E. (1970). Proc. Biochem., 5 (June), 19.
Humphrey, A.E. (1977). Chern. Eng. Prog., May, 85.
Humphrey, A.E. (1978). ACS Symposium Series, 72. (American Chemical Society)
IUPAC-proposal (1982) list of symbols for use in Biotechnology Pure & Appl. Chern.
54,1743.
Johnson D.H., and Berthouex, P.M. (1975a). Biotechnol. Bioeng., 17,557.
Johnson, D.H., and Berthouex, P.M. (1975b). Biotechnol. Bioeng., 17,571.
Kafarow, W.W. (1971). Kybernetische Methoden in der Chemie und Chemischen Tech-
nologie. (K. Hartmann, ed.) Verlag Chemie und Akademie Verlag, Berlin (German)
edition).
Kafarow, W.W., et al. (1979). Modelling of Biochemical Reactors. Moscow: Lesnaja
Promyshlenost. (Russ.)
Kobayashi, T. (1972). Group Training Course, Osaka University, Osaka.
Kossen, N.W.F. (1979). Proc. Soc. Gen. Microbiol. Symp. 29, (Bull A.T. et aI., eds.) 327.
Kossen, N.W.F., and Oosterhuis, N.M.G. (1985). In Rehm, H.J., and G. Reed (eds.).
Biotechnology-A Comprehensive Treatise, Vol. 2. Deerfield Beach, Fla., and Basel:
Verlag Chemie Weinheim, Chap. 24.
Kuhn, H., Friedrich, U., and Fiechter, A. (1979). Europ. J. Appl. Microbiol. Biotechnol.,
6,341.
Mateles, R.I. (1971). Biotechnol. Bioeng., 12,581.
Mateles, R., and Battat, E. (1974). Appl. Microbiol. 105,51.
Mayberry, W.H., et al. (1968). J. Bacteriol., 96,1424.
McCarty P.L. (1970). In Proc. Wastewater Reclamation Reuse Workshop, Lake
Tahoe, Calif., 226.
McLennan, D.G., et al. (1973). Proc. Biochem., 8, 22.
Metz, H. (1975). Chem. Ing. Techn., 43, 60.
Minkevich, I.G., and Eroshin, V.K. (1973). Folia Microbiol., 18,376.
Moletta, R., et al. (1978). Arch. Microbiol., 118, 293.
Monod, J. (1942). Recherches sur la croissance des cultures bacteriennes. Paris:
Hermann & Cie.
Moo-Young, M. (ed.-in-chief) (1985). Comprehensive Biotechnology, Vol. 1-3. Oxford:
Pergamon Press.
Moreira, A.R., van Dedem, G., and Moo-Young, M. (1979). Biotechnol. Bioeng. Symp.,
9,179.
Moser, A. (1977a). Habilitationsschrift, Technical University, Graz, Austria.
Moser, A. (1977b). Chem. Ing. Tech., 49, 612.
Moser, A. (l977c). Chimia, 31,116.

Moser, A. (1978a). In preprints 1st Eur. Congress on Biotechnology, Interlaken,
Switzerland Dechema/D Part 1, p. 88.
Moser, A. (I 978b). Gas-Wasserfach, Wasser/Abwasser, 119,242.
Moser, A. (1980). Paper presented at 6th International Fermentation Symposium,
London, Ontario, July 20-25 (Zajic J.E et aI., eds.) abstract no F.l1.1.7
Moser, A. (1983a). In Adv. in Fermentation 83 (Suppl. Process Biochemistry), p. 202.
Wheatland J. Ltd, England (ed.).
Moser, A. (1983b). In Proceedings of "Biotech 83", International Conference on Com-
mercial Applications and Implications of Biotechnology. London: Online Publica-
tions, p. 961.
Moser, A. (1984). Acta Biotechnol., 4,3.
Moser, A. (1985). In Proc. Advances in Bioreactor Engineering. Lodz, Poland, Septem-
ber 25-27 1985. (Michalski H., et aI., eds.)
Moser, A., and Lafferty, R.M. (1976). Paper presented at 5th Intern. Fermentation
Symposium, Berlin. (H. Dellweg, ed.) Institut fUr Giirungs gewerbe und Biotech-
no logie, Berlin.
Nagai, S. (1979). Adv. Biochem. Eng., 11,49.
Oosterhuis, N.M.G. (1984). Ph.D. thesis, Technical University, Delft, Netherlands.
Ovaskainen, P., Lundell, R., and Laiho, P. (1976). Proc. Biochem., 10 (May), 37.
Papoutsakis, E., and Lim, H.c. (1981). Ind. Eng. Chern. Fundam., 20, 307.
Payne, W.J. (1970). Annu. Rev. Microbiol., 17.
Peters, H. (1976). In Hartmann, L. (ed.). Karlsruher Berichte zur Ingenieurbiologie,
Vol. 9. Univ. Karlsruhe, Inst. t. Ingenieurbiologie und Biotechnologie des Abwassers,
Karlsruhe.
Pickett, A.M., Topiwala, H.H., and Bazin, M.J. (1979). Proc. Biochem., 13 (November),
10.
Pirt, S.J. (1966). Proc. Roy. Soc., (B) 163,224.
Pirt, S.J. (1975). Principles of Microbe and Cell Cultivation, Blackwell Scientific
Publications, Oxford.
Popper, K. (1934,1972). The Logic of Scientific Discovery. London: Hutchinson.
Prigogine, I. (1962). Introduction to Nonequilibrium Thermodynamics. New York:
Wiley-Interscience.
Prochazka, G.J., et al. (1970). J. Bacteriol, 15, 117.
Reardon, K.F., Scheper Th.H., Bailey J.E. (1987). Biotechnol. Prog. 3, Sept, 153.
Rehm, H.J. (1980). Technische Mikrobiologie. Berlin: Springer-Verlag.
Rehm H.J., and Reed, G. (eds.) (1982ff). Biotechnology-A Comprehensive Treatise,
Vol. 1-9. Deerfield Beach, Fla., and Basel: Verlag Chemie Weinheim.
Reuss, M. (1977). Fort. Verfahrenstechnik, 15F, 549.
Reuss, M., et al. (1974). Chern. Ing. Techn., 46, 669.
Reuss, M., et al. (1980). 6th International Fermentation Symposium, London, Ontario.
Richards, J.W. (1968a). Proc. Biochem., 3 (May), 28.
Richards, J.W. (1968b). Proc. Biochem., 3 (June), 56.
Ricica, J. (1969). In Perlman, D. (ed.). Fermentation Advances. New York: Academic
Press, p. 427.
Rock, J.s., et al. (1978). Biotechnol. Bioeng., 20, 1557.
Roels, J.A. (1980a). Biotechnol. Bioeng., 22, 2457.
Roels, J.A. (1980b). Biotechnol. Bioeng., 22, 23.

Roels, J.A., and Kossen, N.W.F. (1978). In Bull, M.J. (ed.). Progress of Industrial
Microbiology, Vol. 14. Amsterdam: Elsevier, p. 95.
Romanovski, J.M., et al. (1974). Kinetische Modelle in der Biophysik. Jena: Fischer.
(German translation by W.A. Knorre and A. Knorre)
Schmid, R., and Sapunov, V.N. (1982). Non-Formal Kinetics, Deerfield Beach, Fla., and
Basel: Verlag Chemie Weinheim.
Schubert, E., and Hofmann, H. (1975). Chern. Ing. Techn.,47, 191.
Shiloach, J., and Bauer, S. (1975). Biotechnol Bioeng., 17,227.
Shuler, M.L., and Domach, M.M. (1982). In (Blanch, H.W., et aI. (eds.). Kinetics
and Thermodynamics in Biological Systems. ACS winter symposium, Boulder,
Colorado. (American Chemical Society)
Shuler, M.L., et al. (1979). Ann. N. Y. Acad. Sci., 326, 35.
Speece, R.E., and McCarty, P.L. (1964). Adv. Water Poll. Res., 2, 305.
Stephanopoulos, G., and San, K.Y. (1984). Biotechnol. Bioeng., 26,1176.
Stouthamer, A.H. (1980). Vierteljahrsschrift der Naturforschenden Gesellschaft in
Zurich, 125/1,43.
Stouthamer, A.H., and Bettenhaussen C.W. (1973). Biochim. Biophys. Acta, 301, 53.
Stouthamer, A.H., van Versefeld H.W. (1987). Trends in Biotechnol., 5,149.
Sukatsch, D.A., and Faust, U. (1977). Proceedings of the Tutzing Symposium, Dechema
Monograph 81, 197.
Swartz, R.W. (1979). Annu. Rep. Ferm. Proc., 3, 75.
Tanner, R.D. (1970). Biotechnol. Bioeng., 12,831.
Tempest, D.W., and Neijssel, O.M. (1975). In Dean, A.C.R., et al. (eds.). Continuous
Culture, Vol. 6. London: SCI, p. 283.
Toda, K. (1981). J. Chern. Tech. Biotechnol., 31, 775.
Topiwala, H.H. (1973). In Norris, J.R., and Ribbons, D.W. (eds.). Methods of Micro-
biology, Vol. 8. London-New York: Academic Press, p. 35.
Topiwala, H., and Sinclair, e.G. (1971). Biotechnol. Bioeng., 13, 1975.
Trilli, A. (1"977). J. Appl. Chern. Biotechnol., 27, 251.
Tsuchiya, H.M., Fredrickson, A.G., and Aris, R. (1966). Adv. Chern. Engng., 6, 125.
Tsuchiya, Y., Nishio, N., and Nagai, S. (1980). Europ. J. Appl. Microbiol. Biotechnol.,
9,211.
van Krevelen, D.W. (1957). In Proceedings ofthe First Symposium on Chemical Reac-
tion Engineering. Oxford: Pergamon Press. (van Krevelen D.W., ed.)
Votruba, J. (1982). Acta Biotechnol., 2,119.
Wang, D.I.e. (1968). Chern. Eng., 75, 99.
Wang, D.I.C., Cooney Ch.L., Demain A.L., Humphrey A.E. and Lilly M. (1979).
"Fermentation and Enzyme Technology" John Wiley & sons, New York.
Wingard, L.B. Jr. (ed.) (1972). Enzyme Engineering. Biotechnol. Bioengng. Symp. 3.
Interscience-Publ., J. Wiley & sons, New York.
Wingard, L.B., Katchalski-Katzir, E., and Goldstein, E. (eds.) (1976). Immobilized
Enzyme Principles. New York: Academic Press.
Zaborsky, O. (1973). Immobilized Enzymes. Cleveland, Ohio: CRS Press.
CHAPTER 3
Bioreactors
3.1 Overview: Industrial Reactors

Bioreactors exist to "tame" biological systems on an industrial scale (see fig. 1.1),
and they should present the optimum conditions for serving the needs of
biological processes. A large number of reactor types are found corresponding
to the large number of different industrial processes (e.g., Moo-Young, 1985;
e.g. Rehm and Reed 1982ff). Reviews with detailed discussions are to be found
in the literature and also in almost all symposium volumes on biotechnology
(Atkinson, 1974; Atkinson and Kossen, 1978; Fiechter, 1978; Ghose and
Mukhopadhyay, 1979; Moo-Young, 1985; Mosen 1985a; Schiigerl, 1979, 1980;
Sittig, 1977; Sittig and Heine, 1977). Here some of the most important types
will be briefly mentioned.
3.1.1 MICROBIOLOGICAL REACTORS (FERMENTERS, CELL TISSUE

CUL TURE VESSELS AND WASTE WATER TREATMENT PLANTS)
A more or less clear overview of the enormous number of very different types
of designs in fermentation, cell tissue culture, food, and waste water and waste
treatment technology is possible. The designs cover technological application,
the stirring and aeration system, and the phase of the main substrates.
The basic types of aeration systems are the submerged sparger, with or
without mechanical stirrer, the surface aerator, and the film reactor. The
simplest type of reactor, a vessel or vat with no special movable stirring or
aeration system, is well suited for carrying out liquid phase, anaerobic cooking-
type operations, such as in brewing. Containers with certain shapes (egg
shaped or cylindrical with tapered top and bottom parts) are preferable
due to improved fluid dynamic characteristics. Vats, together with suitable
horizontal-type reactors such as channels with floating covers, are used in
producing biogas (CH 4 ), in agricultural installations or large-scale community
waste treatment plants. When used on a small scale, such reactors make few
technical demands and may be constructed from fiberglass, reinforced poly-
3.1 Overview: Industrial Reactors 67
ester, concrete, or steel. In this class would also be a rotating horizontal drum
that works like a cement mixer and is suitable for use in fermentations
involving solid substrates (animal wastes, or the fermentation of grain for use
in antibiotic production) (Hesseltine, 1977a,b). Similar designs are used in, for
example, making yogurt (Driessen, Ubbels, and Stadhouders, 1977).
Aerated, stirred vessels can be called standard or reference bioreactors
(Dechema, 1982): Equipped with various stirrers, they are generally suitable
for most uses (Zlokarnik, 1972). Additional mixing of the fluid can be obtained
by building in various barnes and fins. A new development is the so-called
"totally filled bioreactor" (Karrer, 1978; Puhar et aI., 1978).
Multiple stirring devices are advantageous for mixing highly viscous media
such as myceliae fermentations used in producing antibiotics. Column-type
reactors with multiple stirrers and sieve base plates are highly developed
reactors that require a great deal of energy and whose use is justified only in
the case of special problems such as oil-water emulsions. A cascade of stirred
reactors is used, naturally, only for continuous processes. From a technical
viewpoint, the cascade may be thought of as a substitute for a genuine tubular
reactor. The so-called "paddle wheel reactor" is a horizontal container that
has good aeration but can only be built to a limited size (Zlokarnik, 1975).
The self-priming aeration systems using low-pressure air are energy efficient
but are limited to low-viscosity media. A new development is the submerged
tower system with injector aeration ("towerbiology," Leistner et aI., 1979), also
called the "bio-high reactor." Systems that have been used for a comparatively
long time in water purification are oxidation ponds, ditches, and lagoons,
which can be aerated in various ways. New types of systems worth mentioning
are the agitated and aerated tubular reactors (Moser, 1973b, 1977a) and the
scraped-tube reactors (Moo-Young et aI., 1979).
The various types of plug flow bioreactors were recently surveyed by Moser
(1985a). They utilize surface aeration by means of a variety of rotating brushes,
rotors, cone aerators, or gas or fluid jets such as are found in biological waste
water treatment plants. Beyond all mechanically driven systems, reactors can
also be both aerated and mixed pneumatically, or one pump can serve for
both mixing and hydrodynamic stirring.
There is a great deal of interest in bubble columns with their countless
designs: They have no moving parts and are very energy efficient (Deckwer,
1977; Schiigerl, Oels, and Liicke, 1977; Schiigerl et aI., 1978). The air lift
fermenter uses additional fins to achieve the mixing of the liquid (Wang and
Humphrey, 1969). Another new design is the "Andritz" reactor (Paar et ai.
1988). The "pressure cycle fermentor" from Imperial Chemical Industry ICI
(Cow et aI., 1975) and the "deep shaft reactor" (Hines et aI., 1975) also represent
the types of reactors in which the rising column of gas bubbles is used to
circulate the liquid. They have the same characteristics as the loop reactors,
of which there are a large number of variants with internal or external
circulation, most often with jets as the aeration system (Blenke, 1979; the
tubular loop reactor, Ziegler et aI., 1977; the torus or cycle ring reactor,
68 3. Bioreactors
Ui.derach, Widmer, and Einsele, 1978; the cyclone reactor, Dawson, 1974;
the cycle tube cyclone reactor, Liepe et aI., 1978}. The plunging jet reactor
developed by Vogelbusch, Vienna, in collaboration with the Engineering
Center, Bohlen G., East Germany, is another type of special design that works
with a two-phase pump and a foam-like gasjliquid mixture (Schreier, 1975;
Steiner et aI., 1977). The last type of bioreactor to be mentioned is the thin-film
type. Here the liquid and/or solid phase is in the form of a thin layer, and this
promotes the reaction.
Further classifications of biofloc and biofilm bioreactors may include thin-
layer and adhesive fermenters, horizontal trays, the biodisk, and packed bed
and fluidized bed bioreactors (Moser, 1977a). Most reactors of this type are
still used for research at the laboratory scale. Simple trays have long been used
for tissue culture. The fixed bed reactor or percolating filter has long been used
for producing acetic acid, while the biodisk is used in waste water treatment.
Bubble column and air lift reactors are especially suitable for tissue culture
applications, which are generally very sensitive to sheer forces (Katinger,
Scheirer, and Kromen, 1979). Finally, photobioreactors should be mentioned;
they are used to achieve high photosynthetic activity with algae (Jiittner, 1982;
Markl and Vortmeyer, 1973; Pirt, 1980; Pirt et aI., 1983). Integrated reactor
systems using membranes for separation purposes are also being developed.
3.1.2 ENZYME REACTORS

The biologically active enzymes can be used as catalysts either in a soluble,
dispersed form or in a carrier-bound form. Because of the need for the isolation
step and losses of enzyme activity, enzyme processes are sparingly used at
the present time. Immobilized enzymes show promise for minimizing these
activity losses and for facilitating enzyme recovery (Pitcher, 1978).
Usable types of laboratory- or engineering-scale enzyme reactors include
packed bed columns, hollow fiber membrane reactors, and reactors containing
rolled sheets of membrane catalyst. Almost all have some of the characteristics
of plug flow reactors. The so-called "spinning basket" reactor is an analogy
to chemical reactors in which the catalyst is directly attached to the stirring
apparatus (Carberry, 1964). Enzyme reactors currently are being introduced
on a large scale in, for example, the production of amino acids (separation
of d- and I-forms, Chibata et aI., 1972). Surely they will playa larger role in
the future (Coughlin et aI., 1975; Moo-Young, 1985; Mulcahy and La Motta,
1978; Nelbock and Wandrey, 1978; Pye and Wingard, 1974ff; Wandrey, 1983;
Wang et aI., 1979).
The problems of working with carrier-bound enzymes and cells in biofilm
reactors and the biofloc reactors are compared in Table 3.1.
3.1.3 STERILIZERS
Sterilizers are used to kill biological materials. Use of sterilizers is often
necessary to produce infection-free, sterile nutrient media that can be used in
pure cultures.
3.2 Systematics of Bioreactors 69
TABLE 3.1. Biofilm operation compared with floc bioprocessing.

Criterion Floc BiofUm
Mode of operation Discontinuous Continuous (no wash out)
Continuous (wash out)
Process control Multiple, difficult Simple
Product recovery Expensive Easy and inexpensive
Kinetics Homogeneous Heterogeneous
Heterogeneous Resp. pseudo homogeneous
Transport problems Yes Yes
At GIL interface Yes Yes
In L phase Yes Yes
At LIS interface Nearly undetectable Yes (easier to handle)
In S phase Difficult to model Yes
Particle size Size distribution un- Film thickness uncontrolled but possibilities
controlled in STRs for controlled and uniform film growth
From Moser, 1981.
Sterilizers can be operated discontinuously or continuously. For engineering-

scale processes, sterilization is usually done with heat (i.e., steam) on eco-
nomic grounds. Chemical and physical processes also exist (Aiba, Nagai, and
Nishizawa, 1976; Richards, 1968). The stirred processes differ from each other
in method of heat exchange. They are far surpassed for continuous operations
by tube-type sterilizers, a consequence of the formal first-order kinetics of
sterilization (see Chap. 5). Continuous steam sterilizers have been used for
a long time in food processing technology (e.g., milk production).
3.2 Systematics of Bioreactors
A classification scheme for bioreactors can be devised with the help of the
following criteria:
1. The geometry of the reactor, or the way the mass of catalyst is distributed
throughout the reactor volume: biofloc or biofilm reactors. Associated with
this is the question of whether the reactor should be considered a "homo-
geneous" or a "heterogeneous" system, that is, whether it has or does not
move significant physical transport limitations.
2. The mode of operation of the reactor-discontinuous, completely con-
tinuous, semicontinuous, or semidiscontinuous-including transient or
steady-state operation techniques.
3. The state of mixing in the reactor-uniform distribution ("lumped param-
eters") or uneven distribution ("distributed parameters"). Associated with
this is the question of whether the reactor is maximally mixed or totally
segregated, that is, whether the reactor is to be considered an ideal stirred
vessel or an ideal tube reactor.
70 3. Bioreactors
3.2.1 HOMOGENEOUS VERSUS HETEROGENEOUS SYSTEMS
The concept of homogeneous versus heterogeneous used as basis for system-

atization refers to the relationship of the extent or size of the solid mass
(S phase) to the extent of the reaction phase (L phase).
Genuine homogeneous reactions are found only when enzymes in soluble
form are used. The "normal" situation of fermentation technology, in which
biological cells are agglomerated to flocs representing a solid phase is a
heterogeneous system. Nevertheless, such a system can be treated as pseudo-
homogeneous (see Sect. 4.3).
With increasing floc diameter (or in the case ofbiofilm processes increasing
film thickness), the system becomes truly heterogeneous. Transport within and
between different phases is significant, and differential equations must be
formulated for the different situations. These complex cases often can be
reduced to pseudohomogeneous ones by introducing an algebraic factor, and
this is the concept of reaction rate efficiency (cf. Sect. 4.5).
3.2.2 MIXING BEHAVIOR
Process engineering characteristics of reactors deal with the mIxmg con-

ditions of the main reaction phase, the L phase. The degree of segregation
(Danckwerts, 1958) can be used as an example. The two extreme conditions
are referred to as maximal mixing, mm, and total segregation, ts (see Fig. 3.1).
Consider a particular reactor space, for example a tube in which there is
a flow of velocity Vz in the direction z. The entering fluid stream is thought
of as separate layers, and the fate of these layers is observed as they pass
through the reactor space. In the case of mm, there is complete mixing of
the layers over the cross section of the tube. In the case of ts, the layers leave
the reactor unchanged. These extreme conditions are realizable in the reactor
conformations shown in Fig. 3.1c (Zwietering, 1959).
In practice, in the case of stirred reactors, the mixing time tm is used
for characterizing mixing; its experimental measurement is described in
Sect. 3.3.2. That this one-dimensional quantity can represent the actual three-
dimensional time course of mixing is due to the condition of mm, which
occurs at the same rate in all three directions ("lumped parameters" in a stirred
vessel). The state of mixing in reactors with concentration profiles (tubular
reactors or reactors with internal and external circulation) is experimentally
more difficult to observe (Hartung and Hiby, 1973; Hiby, 1972).
F or understanding of the general concept of the degree of mixing in reactors
it is important to recognize that the degree of mixing is made up oftwo equally
important components: (a) micromixing, characterized by the time needed
for mixing in the L phase inside the reactor (stirred reactors), and (b) macro-
mixing, characterized by the residence time distribution (RTD) (continuous
reactors), the reactor being considered a black box.
The two characteristic parameters of micromixing and macromixing, as
3.2 Systematics of Bioreactors 71
CSTR CPFR
c.
In
c c
a
'--_ _ _ _ _ _ _ _ _ t,z z
fIt)
b
JL
-I -I i-v
d
I I I l
JI' ~ ~I { J I I
I I I
mm ts
FIGURE 3.1. Continuous stirred tank reactor (CSTR) versus continuous plug flow
reactor (CPFR). Comparison of (a) concentration profiles, (b) residence time distribution
according to impulse function j(t), representing macromixing, (c) visualization of
micromixing flow behavior with extreme cases of maximum mixedness (mm) and total
segregation (ts) (according to Weinstein and Adler, 1967, Pergamon Journals Ltd.) and
(d) realization of the flow behavior (according to Zwietering, 1959, Pergamon Journals
Ltd.) v, velocity; z, direction.
72 3. Bioreactors
shown in Fig. 3.1, are actually coupled. The simplification implies, however,
that this model using CSTR and CPFR as extreme cases is not appropriate to
adequately represent transition states. The degree of mixing can be especially
well studied in loop reactors, because here the intermediate states appear in
clearer form (Fu et aI., 1971; Lehnert, 1972; Moser, 1985a; Moser and Steiner,
1974, 1975a,b; Rippin, 1967).
Because of the restricted applicability of the concept of mixing time, alter-
natives are sought to characterize mixing more generally. The degree of segre-
gation previously mentioned, and also referred to as the "inhomogeneity," J
(Dankwerts, 1958; Zwietering, 1959), is defined by Equ. 3.1
vara p vara i
J=--=1--- (3.1)
vara var a
and has a range
0c::;Jc::;1
with boundary values
J=O for mm (maximum mixed ness)
J=1 for ts (total segregation)
The quantities var exi' var ex p , and var ex are variances (sum of squares) of the
age distributions at a single "point" (aJ, between two points (a p ), or of the
whole system (ex). The point (Danckwerts, 1958) is defined as an element of
volume that is small in relation to the volume of the reactor but nevertheless
large enough to hold many molecules. At this point the process of mixing,
along with its two extremes, can be conceived of as shown in Fig. 3.1a
and c. The problem with this concept is that the quantity J is-in real
bioprocessing-experimentally almost unmeasurable. It can nevertheless be
used as a model in thinking and working. The cited references demonstrate the
fact that the states of micro mixing and macro mixing are directly dependent
on the recycle flow ratio r of a loop reactor (Dohan and Weinstein, 1973;
Moser and Steiner, 1975a; Rudkin, 1967). The behavior ofloop reactors often
approximates that oftube-type reactors, with total segregation when the value
of the recycling ratio is low. When the recycling ratio is high, the behavior
changes to that of a stirred reactor with maximal mixing (Fig. 3.2). The
consequence is important when, for example, circulating reactors are used to
obtain kinetic data (see Sect. 3.5 and 4.3), and it is also important in obtaining
the optimal configuration and conversion in a reactor (see Chap. 6).
The same model of inhomogeneity has also been used to model conversion
in various combinations of reactors (Tsai et aI., 1969,1971; Wen and Fan, 1975;
see Chap. 6). Finally, a systematic picture of the various types of bioreactors
on the basis of the criteria discussed is shown in Fig. 3.3. Of course, in practice
a strict classification is not possible since the intermediate states (especially
with regard to micro and macromixing) are often dominant.
3.3 Quantification Methods 73
J
ts ... 1
f(vara)
-- --
-- -- ---
0,5
mm . .. 0 L-_L---'-_---'-_----'-_-'-_-'-_-'---_--'-_--'----'
o 0,5
FIGURE 3.2. Relation between degree of segregation J (Danckwerts, 1958) and recycle
ratio r, in a recycle reactor with the variance C( (proportional to BO IOI as a measure of
residence time distribution) as parameter (From Dohan and Weinstein, 1973. With
permission from Ind. Eng. Chern. Fundam., 12, 64. Copyright American Chemical
Society.)
-----------
DISCONTINUOUS
FLOC REACTORS
BIOREACTORS
SEMICONTINUOUS
SEMIDISCONTINUOUS
CONTINUOUS
FILM REACTORS
/~
STIRRED TANK TYPE PLUG FLOW
/~
STIRRED TANK PLUG FLOW TYPE
STIRRED TANKS TUBES,PIPES FLUIDIZED BED FIXED BEDS

BUBBLE COLUMNS TOWERS CMMFF PERCOLATING
CASCADES TUBULAR LOOPS FILTERS
AIR LIFT -R, BIODISKS
LOOP R, LAGOONS TRAYS
DITCHES THIN LAYER R.
CYCLONE R. ADHESIV R.
BARRELS MEMBRANE R.
BASINS
FIGURE 3.3. Classification of bioreactor types on the basis of technical criteria, with
examples of known reactor designs. (From A. Moser, 1977a).
3.3 Quantification Methods

A complete list of all process variables necessary for a detailed characterization
of bioreactors includes the following (Dechema 1984,1987):
- Temperature (T)
- Stirrer speed (n)
- Foam
74 3. Bioreactors
- Gas flow (FG)

- Weight, level (W)
- Addition of feed stock
- O 2 partial pressure (P0 2)
- Redox potential (rH)
- pH
- Sterility
- Biomass (x)
- Gas hold-up (8 G )
- 02/C02 gas exchange
- Power input (P)
- Rheology (v, 1]app' or Kd
- Shear (y, '0)
- Mixing (tm' t c , CTD)
- O 2 transfer (kLl a)
- Morphology (<5* resp. 1]*)
The appropriate methods of measurement and evaluation of the most
essential variables will be briefly discussed here.
3.3.1 RESIDENCE TIME DISTRIBUTIONS (RTD)-MACROMIXING
The experimental measurement and typical results for different residence time
distributions in a continuous reactor are summarized in Fig. 3.4. The same
arrangements used for determining the mixing time are appropriate for deter-
mining the residence time distribution in a reactor. A signal in the form of
a pulse or step function or a periodic function is formed at the input, and
the response is measured at the output.
The shape of the response curve for an impulse function f(t) is shown in
Fig. 3.3b for various residence time distribution functions. Figure 3.4c, shows
the response curves F(t) for a step function. The limiting cases of an ideal
stirred vessel and an ideal tubular reactor are shown in both band c. To
quantify RTD curves, two fundamentally different models are used: (a) the
so-called "one-dimensional dispersion model," primarily used for tubular
reactors with low backmixing, and (2) the so-called "cell model" ("tanks in
series model") which was primarily intended for stirred vessel reactors but is
of general validity.
3.3.1.1 One-Dimensional Dispersion Model (I-parameter model)

The picture involved in the one-dimensional dispersion model is the one-
dimensional process of flow in a tube. There is a flow velocity V z in direction z,
which, in the ideal case, is constant over the reactor cross section aT. Because
of molecular diffusion, turbulent convection, and the parabolic velocity profile
that results from boundary friction (roughness, 8), there are large deviations
from a uniform flow front. The effective longitudinal dispersion coefficient
FIGURE 3.4. Measurement set-up (a) a

for the characterization of the resi-
dence time distribution of continuous
reactors as exemplified by the fluid
phase: (b) pulse method data with the
lIlLC
t t
IMPULSE STEP
ltf-E
pulse function f(t). (c) Step function
data with step function F(t). The ex- b f( t) , - - - - - - - - - - - - ,
treme cases of ideal reactor behavior
is indicated by idCSTR and idCPFR.
t JL-idCPFR
c
F(t) , - - - - - - - - - - - - ,
t _t
DdDeff ) is used as a measure of this effect. It is an analogy to a genuine

diffusion coefficient
(3.2)
Applying Equ. 2.3, the equation for the one-dimensional dispersion model
can be written
ac ac a2c
at = -vzaz + DLaz2 (3.3a)
To facilitate a solution, this equation can be written in dimensionless form

ac/co ac/co ( DL ) a2c/co (3.3b)
at/f = - az/L + V z L a(z/L)2 = 0
In this equation there is a dimensionless number, the so-called Bodenstein

number, Bo (sometimes also called Peelet number, especially when a diameter
is taken as characteristic length):
Bo = vzL (3.4)
DL
The exact solution of Equ. 3.3b, after carefully choice of the appropriate
76 3. Bioreactors
boundary conditions, (see Levenspiel and Smith, 1957) is
fer) = If(t) = Bo .exP [-(1 - r)2 BO ] (3.5)

4nr 4r
In this equation, r = t/f, with I being the mean residence time. The Bodenstein
number is therefore the parameter of the dispersion model used in quantifying
the residence time distribution, and it may be obtained from the experimen-
tally measured curves using Equ. 3.6a with particular boundary conditions
(see Levenspiel and Smith, 1957):
;0 = ~(Jp8~ 1 - 1) (3.6a)
Concerning boundary conditions, it should be stated here that two of the

many possible conditions for a flow vessel are called "closed" or "open"
boundaries, when there is or is not a change of flow, respectively, at the vessel
boundary.
Thus, for "closed" boundaries Equ. 3.6b is used rather than Equ. 3.6a
(3.6b)
and for "open" boundaries
(3.6c)
Because the treatment of end conditions is full of mathematical subtleties

(so that the additivity of variances can become questionable), one should
always state clearly the assumptions about what is happening at the vessel
boundaries before using one of the solutions given in the literature.
In these equations 8 2 /P is the total variance of the distribution function
shown in Fig. 3.4a. It is connected with the spread of the distribution 8 2
(the second moment), which can be obtained directly from the measured values
f(t) versus t when Equ. 3.1 is used with all summations
2 = I t 2 f(t) _ [It f (t)]2

(3.7)
8 If(t) If(t)
The second term on the right side of Equ. 3.7 is the mean residence time, I
(first moment of the distribution, the average value of t).
The numerical value of the Bodenstein number can, in principle, range

between 00 for an ideal tube reactor and for an ideal stirred vessel. One
should remember here that, in the first instance, the dispersion model is
thought of as appropriate for the region of tubular flow. Although the model
may be extrapolated to Bo = 0, one should be very careful in using the
dispersion model when backmixing is great, particularly if systems are not
closed. The use of the cell model is recommended for this region.
The analysis of residence time distribution curves thus far refers to situations
in which the signal is a single pulse. The method of measurement involving

a step function (see Fig. 3.4c) gives curves that are referred to by the symbol
F(t) and, logically, have the following relationship to f(t)
F(t) = J: f(t). dt (3.8)
3.3.1.2 Tanks in Series Model (Cell Model)

Besides the dispersion model, the tanks in series model is the other one-
dimensional model widely used to represent non-ideal flow. Here the fluid is
thought to flow through a series of equal-size ideal stirred tanks, and the
parameter in this model is the number of tanks in the cascade (Neq ). The RTD
curves and moments ofthis model are easy to obtain, since problems of proper
boundary conditions and method of tracer injection and measurement do not
intrude. Neq need not be an integer for curve-fitting purposes. It is strictly
empirical, and no theoretical justification, such as Taylor diffusion, or theoret-
ical estimates of the model parameter, are generally possible. This model starts
from the mass balance equation for a series ofi stirred vessels with 1 sis N,
and N, the number of vessels in the series (or the number of equivalent stages,
N eq )
(3.9)
The general solution for the case of a stirred vessel series with N tanks is,
according to Levenspiel (1972),
NN. t N - 1
f(t) = . e- N t / l (3.1 0)
(N - l)!t N
This equation is simplified in the case of one ideal stirred vessel (N = 1) to
1 /-
f(t) = =' e- t t = D e- Dt (3.11)
t
where D is the rate of dilution that is inversely proportional to t.
Evaluation of the model parameter N is done in a way analogous to Equ. 3.6a
using the variance of the measured distribution 8 2 according to Equ. 3.6d:
p
N=- (3.6d)
82
Recently, an alternative method of estimating N in the range of transition

from plug flows to stirred tank behavior (1 s N s 10) was found with the aid
of a computer simulation of Equ. 3.10. For the given case, the time value of
maximum f(t) value, tmaX' of the individual RTD function in Fig. 3.5 can
be taken as a measure of Neq (Bauer and Moser, 1985). The result of this
computer evaluation is plotted in Fig. 3.5, which can easily be used directly,
to estimate Neq in the range 1 s N s 3
78 3. Bioreactors
N FIGURE 3.5. Computer simulation of

eq
the relationship between equivalent
t 8
number of tanks in series, Neq , and the
time of maximum f(t) function, t max
(see Fig. 3.4b), normalized by mean
I residence time t (Bauer and Moser,

1985).
/
6
/
4
2 /"
~
~
----
o
o 03 06
t max
t
t
N==--- (3.12)
t - t max
There is no exact way to compare the two models (Bo and N), since the RTD
curves are never identical. However, some useful relations are obtained from
equating the variances of the two models (e.g., Kramers and Alberda, 1953):
Bo 1
N eq =2+2: for Bo > 2 (3. 13 a)
or for other special cases (according to Pawlowski, 1962)

Neq = 1 + 1/2JB0 2 +1 (3. 13 b)
This type of comparison of the moments of RTD curves of two models

has wide applicability. For large Neq the RTD curve becomes increasingly
symmetrical and approaches the normal curve of the dispersion model and
a comparison of these two curves allows one to relate the two models. The
range where these conversion equations are valid is determined by the range
of the validity of each of the models.
In evaluating the residence time distributions in a continuous system oper-
ating without recycling, for the case of the ideal discontinuous stirred vessel
the curve with an exponential decay of the dilution process normally appears.
But it appears in such a way that the pulse functions do not overlap (see
Fig. 3.6 and also Blenke, 1979). This means that mixing and dilution processes
are superimposed, and that in reactors that deviate from the ideal continuous
IM~ V_ _ _....J
____ T .. F
~~~_____________~Fr______~~
b f(t)
T
FIGURE 3.6. Measurement set-up (a) for determining the mixing time, t m , and in-
homogeneity, J, for a continuous loop reactor with differing degrees of mixing as
a function of the strength ofthe recycled stream, Fr. (b) CSTRmm behavior with tm < t.
(c) Intermediate case with tm - t. (d) CPFR ts behavior with tm t. (Adapted from
Moser and Steiner, 1975b).
stirred reactor, such as loop reactors, the mixing and residence time distribution
behaviors are manifest in the curves.
The curves obtained experimentally in the case ofloop reactors (cf. Fig.3.6b-
d,) are interpretable as a special case quantitatively using Equ. 3.14 (Moser
and Steiner, 1975a)
i end
f(r) = L fE,i(r)
i=1
= L [(r +1 1)) ( r +r 1)i=1 V~ .exp fBOJ [( - i - tt)2 4t/t
Bo ]
(3.14)
80 3. Bioreactors
The mathematical function describing this type of pulse signal is the sum of
a number, i end , of single functions, fE,i(t), which represent the individual passes
from i = 1 to i end The amplitude is diminished on each pass by a "washing
out" effect in addition to the effect of mixing behavior. The factor preceding
the term containing the Bo number takes into account this dilution effect
(r = Fr/F). The mathematical description of the individual functions uses the
concept of the Bo number, as presented in Equ. 3.5 and Bo E The first and
second moments of the whole distribution function (residence time distri-
bution plus mixing) can be obtained in the same way as Equs. 3.6 and 3.7;
accordingly, this combined function fulfills all the requirements of a normal
distribution function. The mean residence time in the whole system t;ot can be
calculated from
_ V
ttot =p (3.15a)
and the mean residence time of a single pulse function t;nt can be found from
Equ.3.15b '
- V t;ot
tint = F + Fr = 1+r
(3.15b)
The second moment of the whole distribution function is the Bo number,

Bo tot , representing the time distribution in the entire reactor system. In the
-
- - - EQ 3-16
~EXPERIMENT n =100 rpm
50 n =400 rpm
10
o L -_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _~_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _~~
o 5 10
_r
FIGURE 3.7. The longitudinal Bodenstein number of the total pulse function, BO tot
for the liquid phase of a tube-type reactor with recycling as a function of the recycle
ratio, r. The evaluation of the experimental results are compared with the theoretical
calculation, Equ. 3.15b. (From Moser and Steiner, 1974 and 1975).
case of a loop reactor, it is dependent on the recycle ratio, r. As shown in

Fig. 3.7, the plug flow characteristics dominant at the beginning are rapidly
dissipated with an increasingly strong recycle flow F" and the system goes
over to the residence time distribution characteristic of a stirred reactor vessel.
Figure 3.7 shows experimental values (Moser and Steiner, 1974 and 1975a),
and a theoretically derived equation for the variance of the system according
to Fu et aI., (1971):
1 + Nr
(3.16)
tot N(1 + r)
3.3.2 MICROMIXING
3.3.2.1 Mixing Time, tm' and degree of mixing, m

The experimental set-up for the measurement and a typical result are shown
in Fig. 3.8. A measuring device is installed in a reactor. The device is sensitive
to a change of some property: conductivity, pH, color, optical density,
"Schlieren methods," 02' T, radio pill, fluorimetry, radioactivity (Beyeler et aI.,
1981,1983; Bryant and Sadeghzadeh, 1979; Einsele, 1976b; Kappel, 1976;
Kipke, 1984; Middleton, 1979; Moser, 1987; Schneider et al. 1986; Zlokarnik,
1967). The "response function" of such a measurement often has the typical
appearance shown in Fig. 3.8.
The so-called "degree of mixing," m, is sufficient to determine the mixing
time, tm' necessary to reach a particular value of m. Representing the asymptotic
value of the concentration by Coo, the degree of mixing m can be defined with
Equ.3.17
m = c-c .100 00 (3.17)

Coo Co
and it represents the residual deviation from the final concentration in percent.
Usually m is given as 5% or 1%; that is, at a value of 95% or 99% of
the totally mixed concentration. In quoting the mixing time, the degree of
mixing should also always be quoted. Figure 3.9 illustrates a typical plot of
dependence of tm on the degree of mixing.
The mixing time, which was primarily intended as a measurement for
discontinuous stirred reactors, can also be applied in the case ofloop reactors.
As shown in Fig. 3.8b the deviation from Coo can be taken as a direct measure
in the definition of an inhomogeneity, 1 (Lehnert, 1972).
In the case of a continuously operated loop reactor, one can also apply the
concept of mixing time over a wide range. Figure 3.6 brings together the
different results obtainable by superimposing a mixing process such as that
to be found in a discontinuous, closed system on a process of removing fluid
in an open, continuous reactor (Moser and Steiner, 1975a,b). The different
curves in the figure, labeled b through d, result from different ratios, r, of the
82 3. Bioreactors
a
C tm
~t1fi,
t m ,95
t
1,5
b
C
f
~. m
I
0,5
Co 90 95 99
tm - t _m(%)
3.8 3.9
FIGURE 3.8. Measurement (a) and evaluation (b) of the mixing time, tm , of the liquid
phase in a discontinuous stirred tank reactor using the degree-of-mixing parameter m,
Equ. 3.17. In reactors with constant recirculation, the circulation time, t e , may also be
determined as indicated. The degree of inhomogeneity J is also shown. (From Dechema,
1982).
FIGURE 3.9. General form of the dependence of mixing time tm (normalized with tm at
m = 95%) on the degree of mixing m. (From Kipke, 1984).
recycled flow to the output flow (r = Frl F). This occurs simultaneously with
changes in the average value of the time spent in the reactor, t, relative to
the mixing time, t m The envelopes to the curve, indicated by the dotted line,
can be used to evaluate the mixing time, as shown in Fig. 3.6c. Recent work
on structured mixing models using circulation time and CTD is shown in
Sects. 3.3.2.3 and 6.9.4.
3.3.2.2 Mixing Models of Bioreactors (Structured Models)

It should be noted that the RTD character of a whole recycle reactor system,
quantified with Botoo depends not only on the recycle ratio r, but also on
Bo int , the internal RTD characteristic of the reactor (Moser, 1985a).
In practice, situations are even more complicated due to the simultaneous
effect of r on BO tot and Bo int , so that computer simulations fail to describe
experiments as they operate with constant Bo int Experimental evaluation of
BOint from RTD functions shows that the value goes through a maximum at
about r = 10, when using a tubular reactor with recycling (Moser and Steiner,
1975a,b).
Another factor to be considered in analyzing RTD functions on this basis
can be observed by comparing curve band c in Fig. 3.7. Obviously, at Fr F,
the RTD curve approaches that of an ideal CSTR and simultaneously the
micromixing behavior tends in the direction of maximum mixing. Thus, the
degree of segregation J can be expressed as a function of recycle ratio rand
O"~ot and O"fnt' according to Dohan and Weinstein (1973)
r
--1 (2 O"tot +1- r )
--1
J-l
- -
r+ 2
r+ (3.18)
O"int
Especially for technical-scale reactors empirical correlations were developed

(Equs. 3.19 and 3.20 and Tables 3.2 and 3.3) that include the diameter of the
tank dT' or the ratio di/dT. This fact indicates that high-volume vessels are
far from the well mixed (cf. also Sect. 6.9.4). Thus, reactor models must take
into account imperfect mixing; several approaches are used, and they are
summarized in Fig. 3.10.
TABLE 3.2. Correlations between tm and n, resp. d; and dT , for

liquid systems.
Correlation Reference Equ. No.
t m __ n- 1 Kramers and Alberda, 1953 3.19a
I
t m ,...", n- 1 d j 4 Khang and Levenspiel, 1979
van de Vusse, 1955 3.19b
2
tm - (nd j2 )-"3 Norwood and Metzner, 1960 3.19c
2
tm - d~ r
n- (d
I Holmes et aI., 1964 3.19d
2
tm -- n-"3 Biggs, 1963 3.1ge
_ (d j r210-2.S
t -n 1 - Prochazka and Landau, 1961 3.19f
m dT
t -n _ I (djrS/3
- Mersmann et aI., 1976 3.19g
m dT
tm ....... n- 1 to -0.8 Stenberg, 1984 3.19h
tm - n-O. 86 O.07. dT -O.23O.14 Stenberg, 1984 3.19i
From Moser, 1987.
TABLE 3.3. Correlations between tm in case of aerated reactors.

Correction Reference Equ. No.
t m G = 1 + 7.5v O. 27 BG Einsele and Finn, 1980 3.20a

tm
tm = a' n b v~ (0.52 < b < 0.63; c .... 0) Stenberg, 1984 3.20b
tm
d + e' Vs (e -+ 0)
G
-' = Stenberg, 1984 3.20c
tm
From Moser. 1987.

84 3. Bioreactors
a
b
-
c d
----,r
L...--.t_s
FIGURE 3.10. Graphic presentation of mixing models: (a) two-region mixing model,
(b) two-environment mixing model, (c) reversed two-environment mixing model, and
(d) combined backmix-plug flow configuration. In all cases maximum mixed zones
(mm) are combined with zones of total segregation (ts). (Adapted from A. Moser,
1985a, Pergamon Books Ltd.).
A two-region mixing model was successfully employed by Sinclair and

Brown (1970) to explain the experimental deviations observed in a CSTR
(Herbert et aI., 1956). Basically, this model is identical to the so-called "two-
environment model of micromixing" first introduced by Ng and Rippin (1964)
and then successfully applied to bioprocessing in the more empirical reversal
of the original arrangement of mm region and ts environment. The "reversed
two-environment" model (Tsai et aI., 1971) involves an entering region with
mm and a leaving environment of ts. (See, e.g., Diidiikovic, 1977) Similar
attempts have been made by Toda and Dunn (1982) by simulating several
combinations of backmix-plug flow units, representing the flow behavior
ofmm-ts sequences (see Fig. 3.lOd). Brown et aI. (1979) adapted a multiloop
recirculation model previously proposed by van de Vusse (1962), shown in
Fig. 3.11, to quantify imperfect mixing in CSTRs. A series of mixing modules
with feed rates Fj between them are assumed, and unsteady-state mass balances
of each module in sequence are written and solved. This results in a collection
of simultaneous equations relating the performance of each module with that
of its neighbor. Results of varying performance as a consequence of imperfect
mixedness have been described by Sinclair and Brown (1970) and by Toda
and Dunn (1982) (see Fig. 3.12), even though different model approaches
were used. Bajpaj and Reuss (1982) applied the two-environment model to
FIGURE 3.11. Multiloop recirculation mixing model (according to van de Vusse, 1962)
for continuous culture systems. N j , number of mixing modules in series; Vi, volume of
modules; F j , feed rate to one mixing module [m 3 h- 1 ]; F, feed rate to complete system.
(From Brown et aI., 1979).
X/V.SO
~
f(r)
0,5
_M-~,--0.1
o L ____ll.::l=::t:o~'OC5~O,~0~2__.J
0,1 1 10 D/,umax
FIGURE 3.12. Dimensionless cell mass concentration x versus dilution rate D in a

backmix-plug flow configuration (according to Fig. 3.lOd) at varying recycle ratio r.
(From Toda and Dunn, 1982. With permission of John Wiley & Sons, Inc.).
bioprocessing by coupling micromixing and microbial kinetics in the case of

growth of Saccharomyces cerevisiae. They achieved good agreement with
experimental observations when the circulation time distribution model is
used (Bryant, 1977; Mukataka et aI., 1981). This model permits application to
other reactor operation modes, such as batch and fed-batch cultures, in which,
86 3. Bioreactors
YX/s~~--~--~~--'---r-~--~
0,5 t-----_~--.
0,2
0,1 0,2 D
FIGURE 3.13. Yield coefficient YxlS in dependence on dilution rate D in a continuous
stirred tank reactor as a result of modeling the coupling between biokinetics in case
of yeast fermentation and circulation time te' (From Bajpaj and Reuss, 1982).
f (f.ll max ' In 2 )

A 6
B 2
flBotot )
CSTR
2 7
11
FIGURE 3.14. Microbial growth in dependence on micromixing and macromixing:

dimensionless cell mass concentration x as a function of internal recycle ratio r in
a recycle reactor at different values of total Bo number (Bo lol as a measure of macro-
mixing) and varying mean residence time t. (Adapted from Dohan and Weinstein,
1973. With permission from Ind. Eng. Chern. Fundam., 12,64. Copyright American
Chemical Society).
in the case of continuous reactor operation, previously mentioned models

cannot be used. Figure 3.13 is a computer simulation of cell yield Yx1s , which
varies with changing mean circulation time t;, (Bajpaj and Reuss, 1982).
In conclusion, it is clear that a bioreactor needs a high degree of micro-
mixing to operate properly. However, the large amount of macromixing
necessary to allow sufficient micromixing is not favorable to conversion,
because the variable apparent order of reaction is always positive. Clearly
an optimum depends on the coupling between the two components of the
mixing. Figure 3.14 supports this by showing computed cell mass concentra-
tion versus recycle ratio in reactors with different values of var IX (i.e., different
Botot ). Mean residence time is made dimensionless (r) by multiplication with
the factor Jlmax/ln 2. As can be seen, complete conversion cannot be ap-
proached in a CSTR (vafIX = I) for any state of micro mixing, even for infinite
r. Conversion in a reactor with var IX = 0.5 (Bo - 7) increases monotonically
with r until r max is reached. In this case, even at moderate t and at an

intermediate level of segregation, complete conversion can be approached. A
further increase in BO tot (curve 3 with BO tot '" 11) causes the same tendency as
with curve 2 to be exhibited, but a much longer mean residence time is needed
to approach complete conversion because the maximum amount of micro-
mixing permitted is quite small. The optimum value for macromixing and
micromixing found in this simulation study (Dohan and Weinstein, 1973) was
experimentally verified in a CPFR with recycling in the case of beer brewing
and was called "mixed plug flow" (Moser, 1973b,1977b). A greater maximum
production rate of cell mass than that achievable in a CSTR was realized in
a piston-flow loop reactor by theoretical analysis (Grieves et aI., 1964). Similar
experimental findings of an optimum recycle rate, which depends on Bo and
Da (Damkoehler) number are reported for bubble columns (Schiigerl, 1977,
1982) and enzyme reactors (Wandrey and Flaschel. 1979) and represent all-
together a fact quantified ih Chap. 6.
Thus, recycle reactors represent an advantageous device for experimental
verification of mixing models, as emphasized here. Recycle reactors can be
operated in either batch or continuous mode for this purpose, and they also
exhibit the property of short cycle time distribution (CTD), which is very
favorable compared with the CTD of normal stirred tanks.
In practice, in the designing of reactors it will no doubt be advantageous to
use methods utilizing both biological and physical test systems (see Sect. 3.3.11).
It will also be worth testing the extent to which the strategies can be combined.
For example, the relevance of mixing time for bioprocessing seems somehow
clear at first but becomes less clear when the meaning of the degree of mixing
is considered. Einsele et ai. (1978), by examining mixing times and comparing
conventional measurements (method 1) according to Fig. 3.8 with measure-
ments obtained with a fluorometer (method 2), showed that the response time
of biological cells seems to be always the same (4.3 s). It should be remembered
that method 1 contains transport resistances (cf. no. 2, according to Fig. 2.3),
while the culture fluorescence measurements include mass transfer steps 2
through 5, in the legend to Fig. 2.3. Thus, bulk mixing is particularly important
when tm is greater than cell response time of 4.3 s. This is the case in pilot
plants and technical scale reactors. This is further discussed in Sect. 3.3.11 and
in the next section.
3.3.2.3 Experimental Verification of Structured Mixing Models

Additional data are required for the experimental verification of structured
mixing models. The concept of mixing time in liquid phase reactors is widely
used (e.g., Kipke, 1984). The simple method often fails in the case of real
complex and gassed fermentation media, and it also fails in situations in
which the mixing behavior cannot be represented by a single parameter (t m ),
something which occurs in large-scale reactors. Normally problems arise in
the application of basic tracer injection methods such as pH- or conductivity
change: Neither property yields a significant response in real fermentation
88 3. Bioreactors
media due to the media's buffer capacity and high ion concentration. Einsele
and Finn (1980) described a series of experimental results in aerated fermenters
using a pH-electrode. The flow-follower method ("radio pill") was developed
by Bryant and Sadegbzadeh (1979). A flow follower consists of a small en-
capsulated radio transmitter; it has been used by Middleton (1979), Mukataka
et al. (1980), Cooker et al. (1983), and by Oosterhuis (1984). Detection of the
pill outside the impeller area is achieved by mounting two aerials, each
consisting of an isolated steel cable loop, concentrically around the impeller
plane. Limitations in the use of the radio pill are imposed by the relatively
large size of the pill (1-3 cm) in comparison with liquid flow pattern. Con-
ductivity methods have been developed by Stenberg (1984). Disturbances
resulting from gas bubbles passing a conductivity electrode were significantly
reduced with the aid of a series of modified conductivity electrodes (Stenberg
1984). Each electrode was made of 10 platinated wire ends of 0.5 mm diameter.
The 10 wire ends wer't: separated from each other by approximately 1 cm.
The general advantage of conductivity measurements is the small amount of
chemicals needed and the small-scale electrode. Another simple alternative
using a T-method was recently developed (Schneider et aI., 1986). Further-
more, in production-scale bioreactors the cells will always be exposed to a
circulation time distribution (CTD) and not to a mean value of t;,. A typical
plot of CTD would show that the first part of the distribution is better
described than the tail. However, the tail of the distribution is of especially
great importance for microbial processing. A first attempt to scale down a
CTD was made out by Katinger (1976) simulating S concentration in a loop
reactor. Later, Bajpaj and Reuss (1982) coupled mixing and microbial kinetics
using a model of CTD to evaluate the performance of a bioreactor for yeast
production. A number of research workers (Bryant, 1977; Bryant and Sadegh-
zadeh, 1979; Middleton, 1979) measured CTD in stirred tanks and reported
it as being log normal, that is, a two-parameter distribution described by the
first and second moment (f and 0"2) can be used adequately:
f(t)dt = (fiO"l.t)exp[ _~Cnto"~ {t1 YJdt, (3.21)
where f(t)dt is the fraction of circulations whose circulation time lies between
t and t + dt. Here {t1 and 0"1 are the mean and the standard deviation of
normally distributed variables that are related to t and 0"2.
In Figure 3.12 and 3.13 (and later Fig. 3.29) present typical results of the
influence of varied circulation times on bioprocessing. An advanced evaluation
strategy for mixing time, in addition to the methods presented in Sect. 3.3.1,
which contributes to easier verification in experiments, was developed by
Khang and Levenspiel (1979), showing that the liquid mixing process may be
described as a first-order process with a single decay rate constant, k m . If
a salt pulse is added to the system, the concentration at a given point may be
written as
C(X,y,z,t) = C* + f(x,y,z,t)'e- kmt (3.22a)
where f(x, y, z, t) is a bounded function with a mean value equal to zero. At

t = 00 the salt concentration is C = C* everywhere in the reactor. The degree
of homogeneity as a measure for mixing is then described after intergration
over the entire volume by considering the absolute deviation IY.
2.. f C*I :::::: e- kmt 2.. f (3.22b)

a =
N n=l
ICn -
V Jv If(x, y, z, t)1 dV
or a quadratic deviation (variance varIY. or (j2)
(3.22c)
The exponential factors of Equs. 3.22a and b and are assumed to describe the
main time dependence; consequently, the integrals are assumed to be constant
in time. This indicates that plots of In(a) and In((j2) versus t would yield
straight lines, with a slope - k m and - 2k m , respectively. This yields, according
to Equ. 3.22a
e-kmtm~ Iv If(x,y,z,t = tm)ldV = 0.05~ Iv If(x,y,z,t = O)ldV (3.22d)
Since the integral is assumed to be independent of t, the mixing time for

m = 95% is calculated according to
In 0.05
tm'i=-k- (3.23)
m
Mixing time thus can be evaluated in three different ways:

1. By fitting a straight line to In(a) versus t, giving t m l
2. By fitting a straight line to In((j2) versus t, giving t m 2
3. By determining the time for each electrode to lie within 5% from final value,
giving t m 3 according to conventional evaluation (cf. Sect. 3.3.1)
Mixing times evaluated according to IY. and var IY. are in good agreement
with each other and also with the conventional method. Further, it can be
shown that the approximation to a first-order system seems to fit quite well
even when gas is sparged through the vessel.
An alternative method is the regular circulation of the liquid, which Holmes
et aI., 1964, introduced as the circulation time concept tc' Generally, a relation-
ship between tc and tm is valid
(3.24)
where 7 < c < 3 in most cases (Bruxelmane, 1983; Holmes et aI., 1964;
Prochazka and Landau, 1961; Stenberg, 1984). Experimentally, circulation
time can be obtained using the same methods as for tm (Mukataka et aI., 1981),
that is, ionic tracer pulse (conductivity), radio nucleotides (Merz and Vogg,
1978), and the radio pill (see also Reuss and Bajpaj, 1987, and Moser, 1987).
Joshi (1980) calculated the mean t c . L on the basis of tm and the maximum
90 3. Bioreactors
length of circulation loop and found good agreement with experimental data.
Thus, in practice, an exactly reverse procedure can be followed to calculate
tm from tc data. Obviously, tm is directly proportional to the longest loop
length, which depends on the type of impeller and position. Furthermore, tm
is inversely proportional to the mean circulation velocity near the reactor wall
(flat-blade turbine) or near the surface (upflow propeller) (Joshi et aI., 1982)
length of longest loop
(3,.25)
tc = mean circulation velocity
3.3.3 OXYGEN TRANSFER RATE (OTR)

There are several principal methods for measuring the OTR:
1. Sulphite oxidation
2. Gas in/gas out method (physical adsorption)
3. Gas analysis
4. Dynamic methods during the process using P02 electrodes
5. Glucose oxidase method
6. Combined methods using biological test systems
7. New methods (e.g., von Stockar and Stravs, 1983)
The Cu++ - or especially the Co++ -catalyzed oxidation of sulphite (Cooper,
Fernstrom, and Miller, 1944; Reith, 1968) is thought to be suitable for esti-
mating the OTR in comparing and designing gas-liquid (GIL) reactors only
under similar physical conditions as bioprocessing. Mass transport occurring
simultaneously with a chemical reaction is a theoretical problem that results;
it is dealt with in Sect. 4.4.
Fortunately, the purely physical methods used in analyzing input or output
gases can be applied directly in the medium, so that these methods reproduce
the hydrodynamic behavior more accurately than is possible using sulphite
oxidation. Analysis can be done using the laws of physical adsorption. Some
methods, however, cannot be applied directly in the fermentation, during
the run (sulphite and glucose oxidase), or due to problems with the P0 2 elec-
trode (gas in-gas out and dynamic methods using P02 electrodes). Industry
now prefers to operate with gas analyzers because of their superior long-term
stability, and researchers at universities often use P02 electrodes in bench-
scale reactors. Recently, combined methods have been introduced. The basic
methodology will be outlined briefly here.
3.3.3.1 Theory of Methodology

According to the theory of mass transport, the OTR depends on the following
quantities.
1. Specific exchange surface area, a
(3.26)
where GG is the partial volume of the gas phase in the reactor (the so-called
"gas hold-up"). The quantity GG can be calculated from
GG =- =
VG
VR VL
VG
+ VG
=1 - GL (3.27)
and dB is the average diameter of the gas bubbles (the Sauter diameter).
Equation 3.28 can be used to calculate dB from the number of bubbles, n,
each of which has a diameter x
(3.28)
Beyond this physical method using photography or light scattering for

measuring a, the chemical method based on sulfite oxidation is often
successfully used despite disadvantages due to different fluid properties.
According to the theory of mass transfer with simultaneous chemical reac-
tion, A can be calculated from
(3.29)
where no = OTR and 0 = steady-state concentration (cf. Sect. 4.5.4).

2. Mass transport coefficient, kL> a parameter defined and calculated according
to various theories of mass transport, as given in Equ. 3.30. (As O 2 is a gas
with very low solubility, the gas side mass transport coefficient can be
ignored).
Two-film theory, with (j = hypothetical film thickness (as tk -4 00):
D
kL =b (3.30a)
Penetration theory, with tK = GIL contact time
kL = f4D (3.30b)
..Jm:
Surface renewal theory, with s = rate of renewal of surface
kL=~ (3.30c)
Convection theory (King, 1966; Kishinevskii, 1951; Moser, 1973c), with
E = formal convection coefficient [cm2/s]
kL = J(D + E)s (3.30d)
(see also Philipps, 1969):
OTR = kLa(o* - 0) + 21O- 7 as (3.30e)
The convection coefficient takes into account the adsorptive effects of the
gas at the liquid surface; this is particularly important in the case of high
92 3. Bioreactors
a FIOURE 3.15. Experimental set-up

and evaluation of the O 2 transfer
rate characteristics of bioreactors
using the dynamic method for mea-
suring kLl . a. (a) Measuring set-up
b using a step change in inflowing O 2
concentration. (b) Response of the
II III dynamic method. (c) Evaluation
plot. Further explanation is found
in the text.
--t
rates of surface renewal (that is, as tK --+ 0). Normally kL is determined as

kLa (cf. Fig. 3.15) or using sulfite oxidation (see Sect. 4.5.4).
3. Concentration difference, (0* - 0), between the saturation concentration, 0*,
and the actual O 2 concentration, 0 (the "driving force").
3.3.3.2 O 2 Solubility
The saturation concentration is that concentration that is in equilibrium with
the partial pressure of O 2 in the gas phase (Po)
0* = Po (3.31)
HeRT
where He is the dimensionless Henry distribution coefficient (H = He RT).
The problem of O 2 solubility plays a central role in bioprocess engineering
due to its importance for the analysis of O 2 consumption kinetics and OTR,
process control, and scale-up calculations in aerobic processes. Many investi-
gators still use the value for 0* in water as the first approximation because
of a lack of a reliable method. Recently, Schumpe et ai. (1982, 1985) reviewed
the applicable methods: calibration techniques (Baburin et aI., 1981; Kiippeli
and Fiechter, 1981; Lehmann et ai. 1980; Liu et aI., 1973), the classical satura-
tion technique (Popovic, Niebelschiitz, and Reuss, 1979; Quicker et aI., 1981),
and calculation methods based on the van Krevelen-Hooftijzer approach
(1948):
(3.32)
where IJ is the ionic strength of a solution given by

(3.33)
z being the electric charge of the ions. The proportionality factor Ko can be
calculated from the three terms
(3.34)
where i+ = ie.lions
i_ = ianions
Thus, oxygen solubility is not solely dependent on p and T Even this relation
is more complex, as shown by Mihaltz and Hollo (1980) in the following
equation:
(3.35)
Significant differences between actual measurements and tables based on

the CMEA recommendation (1968) initiated the derivation of Equ. 3.35. The
strong influence of salts and of nutrients such as glucose in real fermentation
broths was quantified by the following approach, taking into account the
additive effect of salts and nutrients, Equ. 3.36 (Popovic et aI., 1979):
0* 0* 0*
log* = log-*- + log* (3.36)
0eff 0s.1t 0gl
In this equation, the effect of a glucose solution of concentration cg1 is

considered in terms of a linear equation of the form
(3.37)
Equ. 3.38 gives the effect of the salt solution, which is represented in terms of
the ionic strength as measured by the conductivity A [Q-1 . cm -1 ]
(3.38)
The coefficients ai in Equ. 3.38 are not influenced by the glucose concentration.
Recently an experimental method was presented by Schneider and Moser
(1984, 1985) that can be applied directly to the measurement of 0* in cultiva-
tion media during a process. It is based on a joint analysis of the gas and
94 3. Bioreactors
liquid phase using a paramagnetic analyzer, respectively a mass spectrometer

and a POz electrode, setting both values from gas and liquid analysis equal:
(3.39)
This identity equation gives the value for fo*, from which a:'rr can be calculated
(3.40)
3.3.3.3 Methods for kLl a Measurement

Method 1. Gas in/gas out
The mass transport equation can be written
do
OTR = dt = no = k L a(o:rr - 0) (3.41)
where o:rr represents the effective O 2 saturation and kL . a is the volumetric

mass transport coefficient [h- 1 ], also known as the aeration constant. The
solution to this differential equation is
0* - 0
In--- = -kLat (3.42)
0*
and the value of kL ' a can be found by using the slope of a semilogarithmic
plot. Recently an alternative equation for Equ. 3.41 was proposed (Sinclair,
1984).
Method 2. The dynamic OTR measurement
The dynamic method of OTR measurement is the one most often used in
laboratories (Bandyopadhyay, Humphrey, and Taguchi, 1967; Taguchi and
Humphrey, 1966). It is applicable during processing and uses a sterilizable
POz electrode (Lee and Tsao, 1979). The experimental arrangement is shown
in Fig. 3.15a. A step-type concentration change is created in the fully as-
sembled bioreactor by turning off the air supply at time to and restarting it at
time t l' The response to this signal contains information reflective of the O 2
content. A typical dynamic method response curve is given in Fig. 3.15b.
Incidentally, when t s to there is a equilibrium between OTR and O 2 utiliza-
tion, which results in a constant value of oxygen concentration 000- In phase I
do
dt = kL' a(o* - 0",) - qo' x (3.43)
in the stationary state, do/dt = 0, so that
(3.44)
If O 2 consumption is known (see Equ. 2.5d), one can, in principle, determine

kL . a from a measurement of 000" The accuracy of this type of calculation is
low; therefore one also uses phase II and phase III for this type of calculation.
In phase II, with OTR = 0, Equ. 3.43 may be reduced to Equ. 2.5d, which
describes the rate of respiration of the cell mass. From the slope of the curve,
the parameters of the kinetic model can be obtained when x is known (for
example, qO,max from the maximum slope and Ko from the half maximum;
see Chap. 5).
Differential Equ. 3.43 is completely valid in phase III. With a slight
rearrangement,
= kLl. a (~~ + qo' x) + 0* (3.45)
a graphic presentation is possible, so that with a given value for qo . x one can
obtain the volumetric mass transport coefficient kL . a. This type of graphic
solution is shown in Fig. 3.15c.
An analysis according to this saturation procedure is, however, subject to
many errors (Sobotka et at, 1982); Equ. 3.43 is a simplified representation of
a dynamic process in which, in reality, not only respiration and aeration
processes are at work. Neglected is the additional dependence of the gas phase
dynamics on the separation from, or mixing of, the old or the fresh air or
nitrogen in the reactor volume and the electrode behavior. The dynamic
behavior of both influences can be accounted for by means of a first-order
equation (Dunn and Einsele, 1975). For the gas phase
(3.46)
where kv,G is the dilution rate, or dilution constant, of the gas phase in
the reactor [h- 1 ]. This constant is inversely proportional to tG , the mean
residence time of the gas phase (VG/FG) in cases of maximal mixing
1
kv,G = tG (3.47)
For the electrode response, where the O 2 concentration read is 0E' one has
dOE
at = kE(OL - E) (3.48)
where kE is the time constant of the electrode response [e l ], which very often
is formally of first order (Aiba and Huang, 1969). It is inversely proportional
to the often-used time quantity t E , which is the time necessary for the electrode
response to reach 63.2% of its final value when the electrode is subjected to
a sudden change (e.g., transferring it from a solution with 0L = 0 to 0L = 0*)
(3.49)
96 3. Bioreactors
o
t 1------- ---- ----
1
,STEP
1
1
I
I
1 RESPONSE
I
1
1 EQ 3-50
I
I
I
I
I
-t
FIGURE 3.16. Measurement of kLl . a using the momentum method (Dang et aI., 1977;
Nikolaev et aI., 1976) assuming gas and electrode effects known from the response
curve of the electrode, respectively the aeration system, and assuming a step function
signal.
Available POl electrodes are very slow in responding, and kL . a measure-

ment using the dynamic method results in significant errors. Many of the
methods described in the literature to correct for gas and electrode dynamics
are complicated and require a computer for calculations (Heineken, 1970; Lee
and Tsao, 1979; Sobotka, Linek, and Prokop, 1973; Votruba and Sobotka,
1976). A simple method of evaluation that directly utilizes the response curve
to a step function of the electrode and the aeration system is shown in Fig. 3.16
(Dang, Karrer and Dunn, 1977; Nikolaev et aI., 1976). Use of this relative
procedure allows the influence of the electrode to be eliminated even in viscous
media. The influence of gas phase dynamics can be obtained from the area
between the two curves, Equ. 3.50 (for the case of a maximally mixed gas phase
in a well-stirred reactor vessel)
1 1 VL 1 1
a=-+-'-+--+- (3.50)
kLa He FG kv,G kE
In the case of a gas phase with plug flow, the area between the curves is
inversely proportional to k L ' a (cf. Fig. 3.16). In all cases, fast response elec-
trodes, if available, should be used (kE kL ' a), which minimizes distortions
in measuring OTR.
A simple method of correcting kLa values distorted by electrode lag was
used by Heinzle (1978), following a formal kinetic approach, by a simultaneous
,
k L1 8 1- 010
II E r--------------------,
k L1 8 %
0
10
,."'c;:;:],
0,8
0,1
30
0,5 50
~01~---1----~----L---~
0 0,5 o , , "
k l1 8'/kE t 1/e
3.17 3.18
FIGURE 3.17. Graphic plot of the relation between actual (apparent) value of kLIa,
the electrode response constaht kE' and the true kLl a value. The axis on the right side
indicates the deviation L\ from the true value. The two curves resulting from two
different regions of the evaluation plot (see inserted semilogarithmic plot). (Adapted
from Heinzle, 1978.)
FIGURE 3.18. Semilogarithmic plots of (1 - OE/O*) versus time t of aeration curves at
different conditions (curves B-D) and a step response experiment to determine kE
(curve A). Shifting the curves to intersect the ordinate at unit value yields the time
constants tIle at the 0.37 line, according to Equ. 3.51. (From Ruchti et aI., 1981. With
permission of John Wiley & Sons, Inc.).
solution of Equs. 3.41 and 3.48. The results are plotted in Fig. 3.17, showing
the deviation in percent of actual kLa from the true value kLa with increasing
value of kE. As can be seen, no correction is needed if kLa': kE ::s;; 0.5, and
the measurements are seriously in error if this ratio is ~ 0.5.
Another simplification of kLa calculations was proposed by Ruchti et al.
(1981). This approach again uses the moment method, taking the form offour
first-order lags in series, leading to a simple graphical estimation. The last
dependent variable in the series is delayed by approximately the sum of
the time constants. Thus, for this model, using normalized concentration of
oxygen D, the quantity (1 - DE) attains the value of 1 - e- l (or 0.623) at
approximately
1 1 1
t lle = -k + -k- + -k (3.51)
La V,G E
A further important property is that the (1 - DE) versus t curve will be

approximately exponential for large values of t. This characteristic is evident
in Fig. 3.18, where (1 - DE) from the response data from three experiments
is plotted versus time on a semilogarithmic scale. Also shown in this figure
is the response of the electrode to a step change in G'L' It has been found
empirically that if the lines are shifted but their slope is left unchanged, so that
their intercept with the (1 - DE) axis becomes equal to 1.0, then the kLa values
98 3. Bioreactors
can be calculated from Equ. 3.51. At highest kLa values, in the range of 10 3 h-1,
however, inaccuracies due to the error in tG appear ( 30%) (Bauer and
Moser, 1985).
Finally, there is a method of OTR determination that is especially applicable
in cases of very intensively aerated reactors with high OTR. For this situation,
the dynamic method just mentioned is limited by the electrode's dynamic
response. The so-called "stationary method" (Lucke, Oels, and Schugerl, 1977)
works by passing the liquid phase, which is aerated in the reactor, through
a circuit containing a de-aerator so that a stationary state is reached.
3.3.3.4 Gas Analysis

The basic gas exchange rate is generally correlated with gas balance and can
be calculated by the equation
FG
OTR = (O'G.in - O'G.ex) VL (3.52)
where uG.in and 0G.cx refer to the inlet and exit concentration of O 2 , related
to partial pressures around the reactor. In a steady state of OTR and O 2
uptake rate (OUR) during a bioprocess, the value of kLa can be estimated
from Equ. 3.43 as
(3.53)
where ,10 is the logarithmic mean value of O 2 in the inlet and outlet gas stream,
given by
,10 = (Po, in - pd - (PO. ex - PL) (3.54)

In(PO.in - pd/(Po,ex - PL)
where PL is the partial pressure of O 2 in the liquid phase and PO,in and PO,ex
are the partial pressures in the air stream. In small vessels where perfect mixing
is approached, this averaging treatment is not required. Furthermore, the log
mean value can only be used if the gas exhibits good plug flow and the liquid
phase exhibits high backmixing.
The OUR term, qo' x, can be evaluated from O 2 partial pressures in the
inlet and outlet gas from the relation (e.g., Lehmann et al., 1982)
32 6 (_ _ iiin ) FG,in
qo'X = 224'10 in - Oex1 _ - _ - -V; (3.55)
. 0ex Cex L
with 0, c, and ii being the moles O 2 , CO2 , and N2 per moles of air mixture
(0 = 0.2095, c = 0.0003, and n = 0.7902). Van Meyenburg and Fiechter (1968)
showed that Equ. 3.55 appears because normally the O 2 uptake and the CO2
evolution differ in volume, and a correction of the apparent values is necessary
to account for the differences in flow into and out of the reactor, on the basis
ofthe balance ofN 2 , the usual inert component. Further correction for Tand
p is also needed.
Similarly, the CO 2 formation rate, qc' x, can be calculated from

44.01 6 (_ Hin _ ) FG in
q x=--10 c -c -- (3.56)
c 22.4 ex 1 - Oex - Cex III VL '
Gas phase analysis is superior to P02 electrode methods due to better stability,
and more and more it is being also accepted in fundamental research when
gas flow rate or liquid volume is high enough for accuracy. In analogy to this
approach, a "biological" method for OTR measurement has recently been
recommended (Kappeli and Fiechter, 1980,1981b). In Oz-limited conditions,
the overall maximum OTR of the bioreactor equals the OUR, which is
calculated by Equ. 3.57
(3.57)
For this method it is essential to know the value of ot (cf. previous section).
The use of Trichosporon cutaneum is recommended by Kapelli and Fiechter
as particularly attractive, since this organism is relatively easy to handle,
although weekly subculture from the original strain is needed. It exhibits
a simple and stable metabolism (strictly respiratory strain, even at O 2 limita-
tion; glucose is converted only to biomass and CO 2 , with some formation of
polysaccharides ).
Another new method for the dynamic measurement of kLa uses gas phase
dynamics and consists of continuously measuring the composition of the
outlet gas in response to a step input of a nonreactive tracer such as CO 2
in the inlet gas stream (Andre et ai., 1981). This method is especially useful
under particular conditions for application to high viscosity media and solid-
substrate fermentations.
3.3.4 DEGREE OF O2 UTILIZATION, 1'/0 2

In calculating the effectiveness of an aeration system or the economics of
aeration, important use is made of the degree of oxygen utilization given in %
(3.58)
This can be calculated from a measured balance of the gas phases.
3.3.5 DEGREE OF HINTERLAND, HI

The characteristic number HI, the "hinterland" (Beek, 1969), also designated
by the volumetric number B (Himer and Blenke, 1974) or by the reciprocal
"degree of volume utilization," Iv (Nagel, Kurten, and Sinn, 1972), can be
calculated for a GI L reactor from
HI = ~ = VL = VL ' kL = (1 - CG)kL (3.59)

VL,film At5 AD aD
HI represents the ratio of the whole liquid volume of the reactor, VL , to that
100 3. Bioreactors
volume of fluid found at the gas-liquid exchange surface A in a film of thick-

ness () (see film theory, Equ. 3.30a). The dimensionless number HI characterizes
the size of the gas-liquid exchange surface of an aerator, and it will be
especially significant in the case of so-called "fast reactions" (cf. Sect. 4.5.4).
3.3.6 POWER CONSUMPTION, P
The electrical energy used for mixing and aeration in a reactor is determined
as the quantity of power supplied over a period of time, expressed in kilowatts,
[kW], or [J. S-l]. Power input measurements ofimpellers depend on the scale
of operation (Brown, 1977): On a plant scale, accurate data are obtained for
alternating current from Watt-meter readings
(3.60)
where 11M is motor efficiency and cos ifJ is power factor, which can be measured
directly with a cos ifJ meter or calculated by
A. true power W
cos 'I' = - - - - = - - - (3.61)
apparent power VA
Drive efficiencies 11D and losses of power transmission in the stirrer shaft gland
and bearings occurring as heat to the belts due to excessive slip at the pulleys
are to be taken into account for better accuracy (Brown, 1984). For a given
stirrer speed, it is possible to formulate an equation for the power to the fluid
(3.62)
As an alternative and more accurate method, measurements can be made by
means of torsion dynamometers, which consist of a torsion bar coupling from
which torque can be measured either by surface strain (Aiba et aI., 1973) or
by angular twist. However, relatively expensive equipment must be used, due
to difficulties in interpretation of the torque signal. For laboratory-scale
measurements (P < 1 kW or V < 501), the use of electrical techniques is not
recommended, because circuit losses are greater than the power transmitted
to the fluid. Arrangements are described where the motor is supported by
a piano wire and is free to rotate in air bearings (Calderbank, 1958) or where
the motor is mounted in two radial load bearings located on the motor shaft
(Brown, 1967). Such arrangements are almost friction-free, and they allow
the calculation of power
P = 2nn0 (3.63)
where n is stirrer speed and 0 is radius of torque times the restraining force.
Strain gauges represent another method for measuring agitator shaft power
and are preferred because of greater accuracy (Aiba, Humphrey, and Millis,
1973; Einsele and Fiechter, 1974).
The power input to bioreactors from gas sparging can be calculated from
FIGURE 3.19. Graphic illustration of experimental sit- Po

uation when quantifying power consumption due to
gas sparging, according to Equ. 3.64, respectively Equ.
3.65 (Adapted from Brown, 1977, 1984). For explana-
tion see text.
FO( V~ +RTln-
P=- M O '1'/N-2 Pi) (3.64)
Po Po
The two terms in this equation represent the kinetic energy change of the gas
due to the velocity difference of Vo in the pipe and in the vessel (~vo = VO,pipe)
and the isothermal expansion of the gas from a point in the liquid at the nozzle
(Pi) to a point where the gas leaves the liquid (Po). Useful values for P can be
calculated neglecting the first term. In practice, the kinetic energy term can be
included approximately as a pressure contribution by replacing Po by the
pressure Pm' which must be measured as shown in Fig. 3.19. Thus, Equ. 3.64
becomes
Fo Pm
P=-RTln- (3.65)
Po Po
3.3.7 O2 EFFICIENCY (ECONOMY), EoJ

This term refers to the quotient of the OTR and the power used per volume
E02
OTR[kg
= P/V kWh
02] (3.66)
E02 can also be defined in reciprocal form.
3.3.8 HEAT TRANSFER RATE, Hv TR

Because there is a biological optimum over a relatively narrow temperature
range and because biological processes are exothermic, heat transfer problems
play an important role, even though the amount of heat generated is com-
paratively less than that in chemical processes. An analogy to mass transfer
may be used to formulate an equation and obtain model parameters for the
Hv TR. The experimental set-up for measuring the heat transfer coefficient is
given in Fig. 3.20. The reactor in which the average temperature should be TR
102 3. Bioreactors
HEAT EXCHANGER
T.
'c I':';~ ................. ~w" IT,
Te. in
a
_z
II III
b
-t
FIGURE 3.20. Quantification of heat exchange characteristics. The set-up of measure-
ments contains a counter-current heat exchanger with indicated values oftemperatures
(Tw at warm sides, 7;, at cool sides). The temperature profile shown (a) serves to calculate
the logarithmic mean according to Equ. 3.67 in such real cases of greater linear
extension. From the plot of reactor temperature TR versus time t (b), in analogy to
Fig. 3.1Sb, heat evolution rate r H and heat transfer coefficient kH aH can be calculated
as shown in the text. At time to, the heat exchanger is stopped; and at 1'1 it is reactivated.
is equipped with a temperature sensor and a heat exchanger in which the flow
can be arranged in either a forward or a reverse direction. The appropriate
temperature profiles for a counter-current heat exchanger are shown in part b
of Fig. 3.20. They may be used for the definition and estimation of the
"driving force" for the Hv TR, that is, the temperature difference Ll T. In the
case of a real heat exchanger, where there are long surfaces with different
temperatures, the logarithmic average, Ll T, is used to obtain Ll T. In analogy
to Equ. 3.54, the LlT may be calculated from
fiT = (Tw,ex - Te,in) - (Tw,in - Te.e,) (3,67)

In[(Tw,cx - Tc'in)/(Tw,in - Te,cx)]
As an approximation suitable for small units that can be considered ideal
(that is, without internal temperature gradients and with a constant transport
coefficient) and in equilibrium situation, fi T can be calculated from (Ebert,
1971)
(3,68)
The total heat balance, available on the basis ofthe directly accessible quantity
of the volumetric heat of reaction hv [kJ' 1-1], is made up of several terms,
q [kJ'I- l h- 1 ]
qtot = (~v t t = + (dd~v)r - qTr + qagit - qrad qevap - qGas (3.69)
The vari.ous terms take into account the various "sinks" (q < 0) or "sources"
(q > 0) of the heat balance. These include, specifically, the reaction itself,
transport, agitation, radiation, and heat loss through evaporation and gas
stream. In most cases the last four terms can be neglected (Bronn, 1971;
Cooney, Wang, and Mateles, 1968; Mou and Cooney, 1976). The term for
the reaction, the heat produced, is represented in Equ. 2.4c. The kinetics are,
as always, introduced in the balance equation. All heats of reaction can be
calculated from a series of Tit measurements using Equ. 3.70 when the specific
heat capacity of the reactor system is known
dhv M dT
-=-c- (3.70)
dt V Pdt
This Uhlich approximation uses the value of the specific heat capacity cp
[kJ . kg-I. C- 1 ] multiplied by the mass of the fermentation medium per unit
volume, MIVj [kg/I]. Typical values for fermentation processes are given in
Table 3.4 (Conney et aI., 1968).
The quantity refered to as the volumetric heat of react on hv is unconvention-
ally defined from a thermodynamic viewpoint. The dimensions [kJ '1- 1 ] are
given for practical reasons. Strictly speaking, a (molar) reaction enthalpy has
TABLE 3.4. Typical data of specific heat and heat capacity

with fermentations.
Specific heat, cp Heat capacity, MjV' cp
Medium [kcal' kg-I. C-!] [kcal'l-! . C-!]
Fermentation broth 1 1.01

Glass material 0.2 0.067
Steel material 0.12 0.0094
Total, global value 1.076
From Cooney, Wang, and Maleles, 1968).

104 3. Bioreactors
dimensions [kJ mol- 1 ]. In a modified form this is also used in kinetics as the
specific, metabolic reaction enthalpy with the dimension [kJ . (gL~.x)-l] or
[kJ . (gAS)-l ] (see Sect. S.4.2).
To evaluate the parameter for heat transfer (the heat transfer coefficient, kH ,
[kJ/m 2 . h 0C]), one can begin with Equ. 3.69, which refers to an exothermic
bioprocess; neglecting the last four terms one obtains Equ. 3.71
dh -
-Y=-k
dt H,
.a.AT+r
H H,
(3.71 )
By analogy to the OTR measurement (see Equ. 3.43), the dynamic method
can also be used for heat transfer measurements. The bioreactor is then
treated as an adiabatic calorimeter with constant internal heat production
(Falch, 1968). A typical temperature curve of the dynamic calorimetry is
shown in Fig. 3.20b in analogy to Fig. 3.lSb. At time to, cooling is interrupted,
and the rate of heat production r H, may be directly calculated from the linear
increase in T (phase II). The quantity k H a H is calculated from the slope dT/dt
in phase III, which, in contrast to OTR measurements, is linear. Equation 3.71,
together with Equ. 3.70, is used for this arithmetic calculation. A graphical
method completely analogous to Fig. 3.1Sc is impossible, due to the large
changes in the cooling temperature and the consequent necessity of calculating
a logarithmic average.
An alternative approach for the quantification of Hy TR (which, however,
is difficult to apply to bench-scale reactors, VR < SO 1) follows the concept of
Equ. 3.52. The energy balance on the cooling water is given by
dh y - _
FH 0
-
dt - V2 c
p.H 2 0
(Tex - Tin ) (3.72)
R
Although much of the earlier work involved the use of crude calorimeters,
new techniques require relatively complicated apparatus and procedures for
determining the heat of fermentation. Different methods are cited and are
known as the principle of gradients, the isothermic and adiabatic principles
(in the case of calorimetry), and the quasi-adiabatic method (Bayer and
Fuehrer, 1982; Bronn, 1971; Cardoso-Duarte, et aI., 1975; Eriksson, 1971;
Eriksson and Home, 1973; Forrest, 1972; Imanaka and Aiba, 1976; Jones,
1973; Korjagin et aI., 1978; Luong and Volesky, 1980, 1982; Marison and von
Stockar 1984; Minkevich and Eroshin, 1973; Monk, 1978; Mou and Cooney,
1976; Schauerte, 1981; van Uden, 1971; Volesky et aI., 1982; Wang et aI., 1978).
Finally, Table 3.S presents a summary of the heat transfer coefficient for
a few types of heat exchangers (Bronn, 1971).
In practice, one finds that heat exchange poses engineering difficulties:
Because of the low temperature difference, it is a given that the method of
transferring heat will be somewhat ineffective and uneconomical. Therefore,
attempts are now being made to culture thermophilic microorganisms that
TABLE 3.5. Typical data on heat transfer

capabilities.
Heat transfer coefficient, kH
Type of heat exchanger [kcal m- 2 . h-1. 0C-1] v
External irrigation 500
Double jacket 900
Internal coils 1200
External plates 2000
will grow and produce optimally at higher temperatures, around 50C and
above. Further information on heat formation kinetics is given in Sect. 5.4.2.
3.3.9 CHARACTERISTIC DIAMETER OF THE BIOCATALYTIC

MASS, dp
The concept and the importance of homogeneity and heterogeneity in char-
acterizing bioreactors was mentioned in Sect. 3.2. In choosing the correct
kinetic model, it is important also to include information on possible transport
limitations in the solid phase, for example thick films or large particles.
Measuring this quantity is difficult in an operating plant. Methods involving
a micrometer and phase contrast microscopy are described in the literature
both for particles (Parker, Kaufmann, and Jenkins, 1971) and for films (Korne-
gay and Andrews, 1968). Biofilm operation is characterized in comparison
with floc operation in Table 3.1.
No doubt in all real cases there will be a distribution of diameters d p This
difficulty can be avoided by using Equ. 3.28 to calculate an average diameter
l4 for a particular distribution function (Atkinson and Ur-Rahman, 1979).
l4
The further application of for evaluating the extent of transport-limiting
factors is described in Sect. 4.5.2. Biofilm reactors with mechanical or hydro-
dynamic control of film thickness exhibit a uniform l4
and thus are more
suitable than floc bioreactors for process kinetic analyses (see Sects. 3.6 and 4.3).
3.3.10 COMPARISON OF PROCESS TECHNOLOGY DATA FOR

BIOREACTORS
Table 3.6 summarizes the types of process technology data and bioreactors
thus far described. Establishing empirical correlations between particular
systems generally is done with the aid of dimensionless quantities. The mixing
time, tm' can be made dimensionless by means of Equ. 3.73 (Zlokarnik, 1967)
t .v
NM = ~2 (3.73)
106 3. Bioreactors
TABLE 3.6. Criteria of process engineering characterization of

bioreactors/bioprocessing and range of typical data.
Quantity Symbol Units Numerical range
O 2 transfer OTR kgm- 3 h- 1 0.3-12 (50)
kLia h- I 100-1500 (3000)
kLI m'h- I 0.3-2
aLi m 2 'm- 3 102-10 6
OTR enhancement EOTR ~1
O 2 utilization 1'/02 % 5-40 (90)
O 2 efficiency E02 kg 02' kWh-I 0.2-3.5
kWh'kg02-I 5-0.3
Hinterland HI 103 -20
Power consumption PIV kW'm- 3 0.3-2 (20)
Micromixing (L) tm S 1-200 (500) as !(VR )
Macromixing (L) Bo, Nc CSTR: N = t (Bo -.0)
CPFR: Bo ~ 7 (N ~ 5)
Biological data
Floc size VIA mm 10- 2-5
Film thickness d rom 10- 2-10
Growth: max. rate: J1.max
h- I 0.02-2
Monod constant: Ks mgt- I 2-50 (100)
Respiration QO.max
h- I 0.05-3
,
Yield YXIS 0.05-0.5 (1)
Productivity ri kg'm- 3 'h- 1 0.5-20
Conversion % 0-1
From Moser, 1981.
The power requirement, P or PG, can be transformed into the power number
Np (Zlokarnik, 1967)
P
N = --,;--,-;;- (3.74a)
p p' n
3 S d
or can be referred to the gas flow velocity, F G , (Schiigerl, 1980) as
N' = P/FG (3.74b)

P p(gV)2/3
The aeration rate, F G , can be given as a dimensionless number NA

FG (3.75)
NA = - d3
n'
and the volumetric mass transfer coefficient kL . a can be made dimensionless
using the following equation, which defines the sorption number (Zlokarnik,
1978, 1979)
(3.76a)
or
N~JTR = kL a (;2) 1/3 (3.76b)
In addition to the dimensionless numbers, there are well-known others, such

as the Sherwood (Sh), Reynolds (Re), Schmidt (Sc), Froude (Fr), Bodenstein
(Bo), and Weber (We) numbers. On the basis of these types of dimensionless
numbers, empirical correlations for a large number of bioreactors have been
made (for example, Blanch, 1979; Schiigerl, 1980; Zlokarnik, 1979). The results
of the experimental measurements of process engineering data are often
presented in the form of a graph; they have the form of the relationships
given in Equs. 3.77a and 3.77b. For the volumetric mass transport coefficient
(Ryu and Humphrey, 1972) (see Fig. 3.21)
(3.77a)
For the specific GIL exchange surface, as shown in Fig. 3.22 (Nagel et aI., 1972)
a= a.(~y.v; (3.77b)
where a, /3, y, and w in Equ. 3.77a are experimentally determined coefficients.

For the power consumption of various stirring systems (Blanch, 1979; Rushton,
Costich, and Everett, 1950) (see Fig. 3.23)
(3.78)
k L1 8
51
100 , . . - - - - - - - - - - - - - - ,
FIGURE 3.21. Typical plot of empirical en-

gineering correlation between GIL transfer
coefficient kLl a and specific power con-
sumption P/V as a function of superficial
gas velocity VS,G' The shaded areas (A for
pure water and B for ionic solutions) rep- 1
resent the range of validity of correlations. P/V KW.M-3
108 3. Bioreactors
a
M-1
106
-
Np
10
ST III
FB
V P
MP
10 2
1 10 3 101 10 2 103 10
10
P/V KW.,:;3
NRe
3.22 3.23
FIGURE 3.22. Comparison of GIL reactors of different constructions in a plot of specific

interfacial area a versus specific power consumption P/V: stirred tanks (ST), bubble
columns (BC), jet nozzles (IN), packed beds (PB), venturi washers (V) and plug flow-jet
nozzles (PFJN). (Adapted from Nagel et aI., 1972).
FIGURE 3.23. Typical graphical plot of dimensionless numbers of power consumption

Np versus NRe in case of flat-blade turbine (FBT), paddles (P), and marine propellers
(MP). The laminar transient, and turbulent, regions are indicated (I, II, III). (Adapted
from Rushton et aI., 1950).
and for the power consumption of various stirrers as aeration systems (Ohyama
and Endho, 1955) as illustrated in Fig. 3.24
(3.79)
For the mixing time of various mixing systems (Zlokarnik, 1967) represented
in Fig. 3.25
(3.80)
For the effectiveness of aeration in various bioreactors, see Fig. 3.26 (Zlokarnik,
1978)
(3.81)
and for the special case of surface aeration (Zlokarnik, 1979), see Fig. 3.27
(3.82)
For the mass transfer coefficients, as demonstrated in Fig. 3.28 according to
Sano et al. (1974)
(3.83)
These correlations, among others, quantify and allow comparisons to be made.
They make possible a system of generalizations in which the values from many
systems can be organized. The mathematical functions h in Equ. 3.78 through
Equ. 3.82 also represent correlations for scale-up purposes.
16
10
/P
10
.8 10
.6
.4
.2
1--+---111 ---0-1+-
UN BAFFLED
0
0 2 4 6 8 10
NA LIMIT
3.24 3.25
FIGURE 3.24. Typical plot of power requirement for agitation in gassed reactors,
expressed as the degree of power decrease PG/P versus aeration number NA as a
function of different types of impellers [flat-blade turbines (FBT; vaned disks (V D)
with varying number; paddle (P)J. (Adapted from Ohyama and Endho, 1955.)
FIGURE 3.25. Plot of dimensionless numbers of power consumption N p versus mixing
time NM with the range of different mixing devices with and without baffies [propeller
(I), anchor (II), flat-blade turbine (III), and helical ribbon impeller (IV)]. (Adapted from
Zlokarnik, 1967.)
NOTR
102 -r------------------------------,
1
10
FIGURE 3.26. Typical plot of data showing the relation between oxygen transfer
number NOTR versus modified dimensionless power number N p ' at different conditions.
(Adapted from Zlokarnik, 1979).
110 3. Bioreactors
N N 3/2
p' Fr
FIGURE 3.27. Typical plot of data of surface aerator efficiencies expressed as modified
oxygen transfer number versus Froude number N Fr . The shaded area includes the
range of different tested types. (Adapted from Zlokamik, 1979.)
N N -1/3
Sh~2SC r---------------------------~
FIGURE 3.28. Typical plot of energy correlation for the LIS-mass transfer coefficient
kL2 expressed as Reynolds number based on energy dissipation N Re , respectively
Sherwood and Schmidt number NSh and Nsc (according to Sano et aI.,
1974). The
shaded area includes all engineering ~brrelations known in literature with varied N sc '
while the correlation of Sano et ai. is given with line A.
3.3.11 BIOLOGICAL TEST SYSTEMS
Data about the physical characteristics of a bioreactor can provide only

limited information. The complete characterization of a bioreactor requires
additional studies involving biological test systems ("reference fermentations").
The fluid dynamics and rheological behavior of media are both directly
and indirectly influenced by the presence of biological cells (Fiechter, 1978).
Microbial processes whose growth or production kinetics are specifically
dependent on changes in their medium or the reactor come into consideration
as biological test systems (Karrer, 1978). Because of the central role of mass
yx/sr---------,
PMA/GlL)
0,14 10
0,12
0,10 5
10 50
FIGURE 3.29. Graphic representation of the formal macroapproach to kinetic modeling

of yeast metabolism, showing the dependence of yield coefficient Yx1s , respectively
maximum product concentrlltion PmaX' on initial glucose concentration in batch cul-
tures. Compared are a conventional stirred tank (curve B) and a horizontal tubular
loop reactor with mechanical agitation and aeration (curve A), both working at
constant kLl a-value (700 h -1). (Adapted from A. Moser, 1977a.)
transport, biological test systems are particularly sensitive to micromixing,

OTR, and shearing forces, as indicated in Fig. 2.15.
Naturally, good mixing is a necessary precondition for the mutual contact
of all phases participating in the reaction {see Fig. 2.3}. The adequacy of
mixing is far more important for bioprocesses than for chemical or physical
processes. Bioprocesses not only have a marked optimum but are also often
very sensitive to concentration changes. A good example is bakers' yeast
{Saccharomyces cerevisiae}, which suffers from O 2 repression (02 - Pasteur
effect) and glucose repression {glucose or Crabtree effect}. At glucose concen-
trations exceeding a critical value, Scrit ::::::: 50 mgjI {Fiechter, 1974}, and despite
aerobic conditions, alcohol is produced and cell yield is reduced {see Fig. 3.29}.
In the special case of bakers' yeast, there is, in practice, a special strategy
for the introduction of fresh substrate-the fed-batch culture technique
(Aiba et al., 1976}-that reduces glucose repression by minimizing glucose
concentration. Despite this, when concentrated solutions are added in well-
mixed reactors there are momentary nonuniformities so that the biological
system experiences a change in concentration and a corresponding change
in metabolism {Einsele, 1976a}. Therefore, especially suitable biological test
systems for reactors are those that are sensitive to, for example, glucose or
O2 (Candida tropicalis, Trichosporum cutaneum; Bacillus subtilis, Escherichia
Ecoli, Beauveria tenella). These systems provide more information on the
mixing conditions of bioreactors than do purely physical measurements
{Cleland et al., 1984; Einsele, 1978; Fiechter, 1978; Griot et al., 1986; Karrer,
1978; Kling and Moser, 1986; Moes et al., 1984}. Also, research on the shear
force effects is carried out with the help of a physical {e.g., Grosse et al., 1985}
or a biological test system, Tetrahymena pyriformis {Midler and Finn, 1966}
to study the damage to cells; Pullularia pullulans is a shear-sensitive biological
system. Reuss (1977) presented a quantification of this effect on the basis of
112 3. Bioreactors
investigations by Tanaka et al. (1975a,b), showing that the ratio between

internal nucleotide concentration (cNJ and total concentration (cN) is exponen-
tially dependent on the excretion rate of nucleotides (rN) which is measurable
by measuring extinction (AE 260 )
(3.84)
Excretion rates are correlated to peripheral stirrer speed vtip ' increasing
linearly for Mucor javanicus and Rhizopus javanicus and exponentially for P.
pullulans:
(3.85)
where n is the rotational speed and d i is the diameter of the impeller. A list
of "reference fermentations" was recommended for the practical purpose of
standardization, including the production of bakers' yeast, pentalenolacton
(Streptomyces arenae), xanthan (Xanthomonas campestris), and gluconic acid
with Aspergillus niger (Lehmann et al., 1982; 1985). Recently it has been
emphasized that process kinetic modeling should be incorporated in these
considerations, based on the integrating strategy (Kung and Moser, 1986a;
Moser, 1985a). Meyer (1987) summarized to state of art of biotest systems.
3.4 Operational Modes and Bioreactor Concepts
Reactor operation includes discontinuous and continuous operations, steady-

state and transient-state modes, and a group of intermediate modes that
can be referred to collectively as semicontinuous operation. The modes of
operation are compared in Fig. 3.30.
A discontinuous stirred tank reactor (DCSTR) with ideal mixing conditions
has no concentration profile in space. The process is, however, time dependent
and thus has a concentration/time profile (c/t).
A semicontinuous reactor (SCSTR) operation has most of the characteristics
of a DCSTR in space, and the cit profile is typically as shown in Fig. 3.30
(see also Figs. 3.32 and 3.33).
A continuous stirred tank reactor (CSTR) with ideal mixing has a uniform
concentration both in time and space. As an important consequence, the
concentration at the output (ceJ is the same as that prevailing in the reactor
(c ex = c R )
The ideal continuous plug flow reactor (CPFR) has no profile at any point
of the tube in the steady state. The process, however, advances along the tube,
and so shows a longitudinal concentration variation. The profile of a CPFR
in space is identical to the profile of a DCSTR in time in case of a constant
volume process; this fact is of great importance for process design ("kinetic
similarity").
Finally, Fig. 3.30 shows the behavior of a cascade of CSTRs with a number,
3.4 Operational Modes and Bioreactor Concepts 113
BASIC
CONCENTRATION PROFILES
REACTOR TYPES
(HOf10GENEOUS) cit C!Z
t!J :~
------- to
tex
dCSTR tex
11 1213
~~ ~
sCSTR
~ CSTR
cex
-I
Co
CPFR
t--
z=L cex
z=o
tCex Co ---------
z=L
:A z=L
1 2 1
I I ~
~~l6J ~
NCSTR
FIGURE 3.30. Basic reactor concept and concentration-versus-time and concentration-

versus-space profiles. DCSTR, discontinuous stirred tank reactor; SCSTR, semicon-
tinuous stirred tank reactor; CSTR, continuous stirred tank reactor; CPFR, continuous
plug flow reactor; NCSTR, a cascade of N stirred vessels.
N, of units in a series (NCSTR). As with the CSTR mode of operation, the

concentration profile in each tank is uniform in both space and time. However,
over the entire length of the cascade, the space/concentration profile will show
a typical step function curve. It is not difficult to see that this spatial behavior
approximates that of a CPFR. As a rule, a cascade with N ~ 5 can actually
be used as a process engineering substitute for a CPFR (see residence time
distribution, Sect. 3.3.1).
The operational modes of bioprocess technology, especially the semi-
continuous modes, can be categorized using a function describing the time
dependence of the input. This is done in Fig. 3.31. The discontinuous and fully
continuous processes are special cases with Fin = 0 and Fin = FeX' respectively
through a constant reactor volume. In the case of a CSTR there are practical
difficulties; a constant, continuous input is not particularly easy to arrange.
114 3. Bioreactors
F F(t)
0
Feed rate function
/~ Fex 0 0 Fex 1- 0
Fin 0 0
/~ /~ Fin 0 F(t) Fex 1- Fin Fex
DISCONTINUOUS DISCONTINUOUS CONTINUOUS CYCLIC or CONTINUOUS

CONST. VOLUME FED BATCH FED BATCH PERIODIC CONST.VOLUME
DCSTR
/l~ CSTR
REPEATED REPEATED REPEATED

BATCH OJ SCONTINUOUS CONTINUOUS
FED BATCH FED BATCH
FIGURE 3.31. Systematic, engineering classification offermentation processes based on

the behavior in time of the input stream, Fin, and the exist stream, Fex.
DISCONTINUOUS FED BATCH CONTI NUOUS FED BATCH
JLJL L'
3
Fit
V/t )Cl
3:
IVY
Sft 3
~2
'. . .
: . , .
. : : :
QUAS I - STEADY STATE 1
, . . . "EXTENDED CULTURE"
FIGURE 3.32. Discontinuous and continuous feed systems characterized by the time
dependence of the input, F; the volume of the liquid phase, V; and the substrate
concentration, s. The continuous feeding types are designated 1~ 3.

The processes with Fex = can be further subdivided into discontinuous
input and continuous input processes.
Among the cyclic or periodic processes with Fin =1= Fex' but Fex =1= 0, one can
distinguish cyclic processes with discontinuous or continuous input from
simple, cyclic discontinuous processes. The time dependence of the addition
ofliquid, Fit, the volume ofliquid in the reactor VL , and a substrate concentra-
tion s are shown in Fig. 3.32 for substrate feeding operations in a fermentation
process. They are shown in Fig. 3.33 for cyclic fermentation processes. The
following simple equations are suitable for a mathematical representation of
continuous feed operations.
(3.86a)
or
REPEATED REPEATED REPEATED

BATCH DI SCONTINUOUS CONTINUOUS
FED BATCH FED BATCH
JL'
to
F,lt ~ ~
U
v- J 1/Lv
V/t ~ ~
~ :
..
Sit
~I\ \f\ IV
. .
..
FIGURE 3.33. Periodic, cyclic fermentation processes characterized by the time de-
)endence of the input, F; the volume of the liquid phase, V; and the substrate
;oncentration, s. The arrow designates the moment when each cycle of the process
Jegins.
116 3. Bioreactors
PROCESS LIFE CYCLE

1 REGIME
2
RELAXED
STEADY- 'fe'fr
STATE
3
PERIODIC
OPERATIO 'fe'" 'fr
4
QUASl-
STEADY 'fe>'fr
FIGURE 3.34. The four classes of periodic reactor operation as defined by Bailey (1973)
represented by typical reactor responses (r) at different environmental changes (e) and
characterized by the ratio of characteristic times 'tr' respectively 't., as indicated.
(3.86b)
The so-called "extended culture," which operates with constant substrate
concentration, is a special case of a mode of operation with constant feed.
The criteria of the classification system recommended by Bailey (1973)
involve a comparison of the time delay between a signal, for example, 't e, and
its response, f r. As shown in Fig. 3.34, transient operation techniques consist
of true intermediate periodic operation (fe '" 't r), and relaxed steady-state
("frozen" systems with fe 't r), and quasi-steady-state ("balanced" systems
with'te fr) operations. With this system a great many bioprocesses fall into
the category of cyclic continuous feed operations (Pickett, Topiwala, and
Bazin, 1979).
The principal advantages ofthis type of cyclic system with transient operating
techniques are apparent in bioprocesses whose maximum productivity is in
a transient region. The products of secondary metabolism (Pirt, 1974) are a
typical example of this group of processes. Another group consists of processes
whose optimal operation requires an optimal substrate concentration-
biomass production with bakers' yeast, for example (Aiba et aI., 1976)-or
where the process is subject to substrate inhibition. An important area of
application for this is in biological waste water purification. These periodic
modes of operation generally show increased productivity. More systematic
and detailed study is needed in this area.
FIXED FLUIDIZED SPOUTED

BED BED BED ELUTRIATION
FLUID FLUID
~P
t
APmf
FLUIDIZED BED
I
SPOUTED BED
v
ms Velutr - vfluid
FIGURE 3.35. Differing situations in a fluid/solid reactor as a function of the velocity

of the fluid phase, VfIuid' along with fluidization diagrams for the various reactor types.
The pressure drop, I1p, increases in solid bed operations until the minimum velocity
for fluidization is reached, Vmf Above this velocity, fluidized bed conditions exist. In
case of particles with diameters about 0.5 cm, the spouted bed can be operated in
fluidized conditions above the minimum velocity of spouting (vms). A further increase
in velocity finally leads to the elutriation of the solid phase.
Finally, the concept of the fluidized bed reactor should be briefly mentioned.
Like so many others, this process technology stems from chemical reactor
technology, and it has its roots in solid bed reactors. In the later, one can
introduce a gas or liquid stream from below that can "fluidize" the particles
as the velocity of the stream, Vfluid' increases.
The situation is diagrammed in Fig. 3.35. Above a particular flow velocity
the small solid-phase particles become suspended. This point is called the
"minimum fluidization velocity," Vrn [, and it occurs at the point where the
pressure drop in the column, IIp, remains constant. This working range is that
of the "fluidized bed." With further increases in Vfluid' the solid phase would
be prematurely "washed out" (elutriation). Another fluidization technology
available for particles with diameters of about 5 mm is called a "spouted" or
"whirling bed" reactor (Mathur and Epstein, 1974). A true fluidized bed has
a porous plate at the bottom of the column. In the whirling bed mode of
operation, the fluid is introduced directly through a small opening in the
118 3. Bioreactors
bottom of a cone-shaped column. A reactor type representing a cross between

a fluidized and a spouted bed was developed for three-phase (GLS) fluidiza-
tion, and it is called a "spouted fluidized bed" (Kono, 1980).
Because of the hydrodynamic effects in the fluidized reactor, mixing of
all the phases involved (gas-liquid or gas-liquid-solid) is improved. The
advantages of the fluidized bed are, among other things, improved mass
and heat transfer coefficients, increased interfacial area, and high biomass
concentration (x > 80 g/l).
The fluidized bed concept is now beginning to be applied in biological waste
water treatment (e.g., Atkinson, 1980) and in food technology (e.g., Moser,
Kosaric and Margaritis, 1980). Enzyme technology has already successfully
operated with fluidized beds (Coughlin et aI., 1975) even though, on a larger
scale (even with chemical processes), all the problems are not yet solved
(Andrews, 1982; Baker et aI., 1980; Werther, 1977, 1980).
3.5 Bioreactor Models

There are three basic homogeneous reactor models (DCSTR, CSTR, and
CPFR, Fig. 3.30) that can be considered ideal cases for calculating conversion.
The equations for balancing all reactor models derive from the conservation
of mass equation, Equ. 2.3. The equation for a balancing all types of ideal
continuous stirred tank reactors (idCSTR) (c ex = CR) can uniformly be based
on a consideration of the following (see Fig. 3.36):
Total change in mass in VL = input mass with Fin - output mass with Fex
change in mass due to reaction (3.87a)
or, in mathematical form,
(3.87b)
~n Pin
X in v
Sin
FIGURE 3.36. Derivation of mass balance

equations that can be based on an ideal
dCSTR stirred tank (condition Cex = cR , where CR
is the actual concentration in the vessel).
Fin = Fex All three configurations are included
sCSTR Fin=F(t) (DCSTR, CSTR, and SCSTR).
3.5 Bioreactor Models 119
3.5.1 MODEL 1: THE IDEAL DISCONTINUOUS STIRRED

TANK REACTOR (DCSTR)
The special conservation equation for the DCSTR, assumed to operate with
ideal mixing and therefore with div grad ni = 0 (see Equ. 2.3), follows from
Equ. 3.87b, where Fin = 0 = Fex and V = constant, that is, dV/dt = 0:
dc
-'=
dt +r
-,
(2.3d)
This equation is not only the conservation of mass equation; it is also the
definition of the rate of formation (ri > 0) or consumption (ri < 0) of a com-
ponent. To calculate conversion, one integrates this equation.
3.5.2 MODEL 2: THE IDEAL CONTINUOUS STIRRED TANK

REACTOR (CSTR) WITH V = CONSTANT
Under the conditions Fex = Fin = F with V = constant and Cex = CR' the mass
equation is
dc
V dt' = F(c in - cex ) riO V (3.87c)
If one uses the quantity t [h] = mean residence time or the quantity D
[h- 1 ] = dilution rate of the fluid phase (cf. Equ. 3.47)
(3.88)
one obtains from Equ. 3.87c

dc
dt' = D(Cin - cex ) ri (3.89)
In the steady state (dcddt = 0), one obtains the following mass equation for
the CSTR
(3.90)
If, for example, one is dealing with a microbial growth process in a CSTR,
the Xin = 0, and one may apply Equ. 2.5a for the rate of growth and thereby
obtain the well-known relations for the equilibrium state. This is the basis
for the CSTR, the so-called "chemostat" or "realstat" mode of operation of
a continuous fermentation process (see Sect. 6.1).
D = Jl (3.91)
3.5.3 MODEL 3: THE IDEAL SEMICONTINUOUS STIRRED

TANK REACTOR (SCSTR) WITH V = VARIABLE
In this case, dV/dt "# 0 (see Fig. 3.36) and therefore Equ. 3.87b remains entirely
valid. For the calculation of a continuous feed operation as shown in Fig. 3.33,
120 3. Bioreactors
Fin is a constant and Fex is zero. In the case of an additional feed according to

Equ. 3.86a, with <Xl = but Pl > 0, then Equ. 3.87b can be rewritten to arrive
at Equ. 3.92 as the mass conservation equation for the SCSTR with variable
volume
(3.92a)
or
(3.92b)
For the case of a growth process, using Equ. 2.5a for the rate of growth and
with Xin = 0, one obtains (Dunn and Mor, 1975)
(3.93)
This equation describes the "quasi-steady state" in which this type of addition
process can be maintained for some time (D = fl.), as shown in Fig. 3.37 with
simple Monod kinetics.
Equation 3.93 can be compared with Equ. 3.91 for the true steady state in
a CSTR. The two are formally identical. The difference is in the real meaning
of the factor D . Cex; in Equ. 3.90 it represents a dilution effect in the case of a
chemostat. In the case of a semicontinuous stirred vessel with variable volume,
the factor D . X R in Equ. 3.93 describes the decrease in cell concentration due
to the change of the volume in the reactor (Dunn and Mor, 1975). Similar
equations can be formulated for all other types of fed-batch processes using
the appropriate modification of Equ. 3.86.
FIGURE 3.37. Computer simulation of fed-batch pro-

cesses: approach to and attainment of the quasi-steady
state with D = Jl from the dimensionless kinetic model
(see Equs. 3.92 and 3.93). The dimensionless parameters
are D = Fin/V JlmaX' X = X/Yo Sin' ji = Jl/JlmaX' and s =
S/Sin. (From Dunn and Mor, 1975. With permission of
John Wiley & Sons, Inc.).
3.5.4 MODEL 4: THE IDEAL CONTINUOUS PLUG FLOW

REACTOR (CPFR) OR TUBULAR REACTOR
To quantify the balances in a CPFR, one may begin from the analogy between
the space profile in a CPFR and the cit profile in a DCSTR (Fig. 3.30). This
implies a reformulation of Equ. 2.3d so that the time and space coordinates
are interchanged. Dimensional analysis shows that the term representing
length dependence must be multiplied with a velocity. Thus, for an ideal
CPFR, with z is the length coordinate, the conservation equation is
dC j
U
Z
- = +r
' dz - I (3.94)
In this equation, Uz is the apparent flow velocity [mh- 1 ], obtained from

the mass flow rate F [m 3 h -1] divided by the cross-sectional area of the tube
A [m 2 ].
The same result is found by proceeding in the usual way, developing a
differential conservation of mass equation for each volume element in the
CPFR, as diagrammed in Fig. 3.38. This figure shows recycling in an amount
Fr [m 3h -1] for maintaining cell culture, so that the recycling ratio is r = Fri F.
For the stationary state and with dV = A dz, the equations for the conserva-
tion of cell mass, x, and substrate, s, are
F(1 + r)' ds + rs' A- dz = 0 (3.95a)

F(1 + r)' dx - rx ' A . dz = 0 (3.95b)
These two equations, along with the equations for the boundary conditions,
form a system of nonlinear differential equations that are very difficult to solve
in closed form; recourse is therefore made to computer techniques (numerical
methods). With certain simplifications, direct integration is possible. The
appropriate boundary conditions are found in this case by considering the
balance of mass at the point M where entering fresh medium meets with
the recycled medium (see Fig. 3.38). For z = 0, one then has
,dV,
,, ,,
S ' :SdS F
-' ,- v
X : ,X.dX z
, : X,S
-'dz'- _ z
FIGURE 3.38. Derivation of mass balance equations in the case of an ideal tube reactor.
Here, differential mass conservation d V = A . dz for s and x for a tubular reactor with
recycle, leading to Equ. 3.85 at steady state.
122 3. Bioreactors
So + r sr (3.96a)
Si =
1+r
and
Xo + r Xr
Xi = (3.96b)
1+r
From Equ. 3.95 and using the relationship dV = A . dz, one can verify the
validity of Equ. 3.94 for the case of no recycling.
The mass balance equation for the case of a cascade of N continuous stirred
reactors, or for processing variants involving the return of concentrated solid-
phase material, can be formulated with analogous considerations (Powell and
Lowe, 1964).
3.5.5 MODEL 5: THE REAL PLUG FLOW REACTOR CPFR

WITH DISPERSION
The quantification of real reactors, those with deviations from ideal plug flow,
can now also be undertaken. In agreement with Equ. 3.3a, in addition to
the convection term operating via the flow velocity Vz> the term with the
dispersion coefficient DL is also operative.
The conservation of mass equation for this situation, which is directly
applied in modeling tube reactors (F. Moser, 1977) and bubble columns
(Reuss, 1976), is thus identical to Equ. 3.3a. These types of one-dimensional
one-phase models are not only necessary for calculating conversion: They are
also very useful in, for example, calculating the kL a value of a reactor with
a concentration profile:
dOL * dOL d 2 0L
dt = 0 = kLa(OL - od - vLh + (1 - GG)DL dz 2 (3.97)
In the steady state, for the actual determination of kL a one needs not only
the experimental measurement of the O 2 concentration along the length of
the reactor; one also needs to know the dispersion coefficient, which can be
obtained from Equ. 3.4 in Sect. 3.3.1, and, for stationary method, Sect. 3.3.3,
Lucke, Oels and Schugerl, 1977).
The types of models described thus far have already been characterized as
one-dimensional one-phase models. This means that only a single coordinate
(the z axis) is considered; in reality, in the case of, for example, flow in
a tube, processes occurring in a radial direction (coordinate R) can also be of
importance.
The formulation of a two-dimensional one-phase model for dispersion takes
the following form, according to Equ. 3.3a:
oCi = -v oCi D 02Ci DR ~(R OCi) (3.98)

ot z OZ + z OZ2 + R oR oR
In addition to having the dispersion coefficient Dv this equation also has

a term for the radial dispersion coefficient, DR, which must also be measured
to refine the model.
3.5.6 THE DISCONTINUOUS RECYCLE REACTOR (DCRR)

An adequate experimental device to avoid serious limitations in the sample
size needed to analyze relatively slow reactions such as fermentation processes
is illustrated in Fig. 3.39. This is the DCRR. The operating conditions are as
follows:
- Differential operation in VR
- Good mixing at high recirculation
- Accurate measurement of C i in adequate time intervals
As a consequence, the concentration will be uniform. Hence Vn; = 0, and thus,
f
from Equ. 2.3
f v
dc
_ 1 dv =
dt v
'jdv (3.99)
Due to the differential performance of VR , the result is

dC i (lI;. + V~)
,.=_._--- (3.100)
1 dt VR
Therefore, a high ratio of lI;./VR reduces the problem of correcting the equation
for the definition of reaction rates,* but at the same time it may prolong the
time needed to achieve measurable conversion.
3.5.7 THE CONTINUOUS RECYCLE REACTOR (CRR)

Figure 3.40 shows a schematic diagram of a continuous recycle reactor, CRR.
With high values of Fn the whole reactor (balance line 2) approximates
a CSTR, working at differential conversions with 'i,2 according to Equ. 3.90.
Writing the balance for line 1, the equation for r i 1 is
F + Fr
'i,l = -v.--(c ex - Cin) (3.101)
R
Since 'i, 1 = 'i,2, it follows that
(3.102)
Thus, if Fr is sufficiently high (Fr F), as was first assumed, C 1 ~ Cex and
* In the case of high sample volumes, it should be kept in mind that at each recycle
the reactor behaves as a steady-state CPFR. Connecting tubes between reactor VR and
recycle reservoir V, should be as short as possible.
124 3. Bioreactors
1------------ - ------1
I 1--------;-] :
1 : I V I 11 I
J -rc
1 R 1
1
1 F+f'i'C1 I
r L _______ J ex
1
1
1-- -- ---------- - - - - - 1
1 1
1 1 1 r--------
I
11 :
I V I ..-
--
1 1
1
1 1 1
1
1 1 1
1 R I
1
I
1
1 I J I
L _________ I
1
I
I
1 I
--
1
F,. 1
..-
~
~
I
1 ~
I F,.
1
1
1
1
I
1
Vr -- c :>
-4~m------------~-i~x L __________________2~
I
F F
3.39 3.40
FIGURE 3.39. Derivation of mass balance equations based on the concept of closed-loop
reactors. The total liquid volume in the recycle, J-;. (reservoir) is recirculated with
the aid of a pump with a flow rate Fr through the reactor volume VR The dotted
lines indicate the balance lines of the reactor (1) and the whole system (2).
FIGURE 3.40. Derivation of mass balance equations based on the concept of open loop
reactors with a stream F through the whole system (balance line 2) and a recycle flow
rate Fr. The concentration in the reactor inlet (cd is different from Cin due to recycle.
under such conditions the global reactor system of a CRR can be treated as
an ideal CSTR, according to Equ. 3.90. Logically, the condition Fr F or
r ~ 100 is only an overall formulation, which can be written more accurately
by considering the dependence on the reaction rate (Luft and Herbertz, 1969).
Using a CRR is important mainly in the case of GIS or LIS catalytic reactions
(e.g., Paspek et aI., 1980) and in the case of reactions with immobilized cells
or enzymes (e.g. Ford et aI., 1972).
3.5.8 MULTIPLE PHASE BIOREACTOR MODELS
Figure 3.41 classifies the various types of bioreactor models according to the
criteria floc-film, stationary-nonstationary, homogeneous-heterogeneous,
and gradient-free-gradient bioprocessing. The types of models presented thus
far are understood as single-phase reactors that consider the liquid phase and
operate under pseudohomogeneous conditions. Section 4.3 deals with tests
for whether and under what conditions, the pseudohomogeneous model is
valid. Because gas, liquid, and solid phases are all present in bioreactors,
3.5 Bioreactor Models l25
BIOREACTOR MODELS
~~
FLOC REACTOR
FILM REACTOR
~~
STEADY STATE
NON STEADY STATE
OPERATION OPERATION
HOMOGENEOUS
/~ HETEROGENEOUS
/~
TRUE
BALANCED
SYSTEMS SYSTEMS DYNAMIC OR
<l-PHASE-, FROZEN
L-PHASE-
REACTORS)
K'~ SYSTEMS
/~
GRAD I ENTLESS
REACTORS
2-PHASE-
REACTORS
3-PHASE-
REACTORS
REACTORS WITH G/L-, G/L/s-
(CSTR,RR) GRADIENTS L/S-
(DCSTR,CPFR) REACTORS REACTORS
FIGURE 3.4l. Classification of mathematical models of bioreactors according to en-

gineering criteria. (Adapted from A. Moser, 1977a)
heterogeneous models in two or three phases should be more realistic. Due

to their complexity, however, such models are only beginning to be applied.
One such two-phase dispersion model has contributed to the successful clarifi-
cation of oxygen limitation and inhibition in the case of a tall bubble column
reactor. The two-phase model considers the gas and liquid phases, and it
involves both mass conservation equations (Reuss, 1976).
In addition to the equation for the liquid phase (according to Equ. 3.97),
which needs only extension with one reaction term, namely -(1 - GG)ro ,
the equation for the gas phase in the steady case can be written
(3.103)
VSG is the superficial gas velocity [m h -1] and is, along with at, dependent
on the hydrostatic pressure in the column.
Some insight into possible types of fluid flow of the reaction phase was given
in Fig. 3.10. Both the liquid and gas phases can exhibit all intermediate stages
and combinations of the extremes of maximum mixing and total segregation.
Close observation of the situation in a reactor to be simulated provides the
first hint as to which model to use.
Performance models of biofilm reactors (fixed and fluidized bed) following
the concepts of pseudo homogeneous and true heterogeneous modeling are
outlined in Sect. 6.7.
126 3. Bioreactors
3.6 "Perfect Bioreactors" in Bench and Pilot Scale

for Process Kinetic Analyses
Bioreactors must fulfill a twofold purpose in bioprocessing-industrial pro-
duction and process kinetic analysis. A standard research bioreactor has
been recommended by the European Federation of Biotechnology (Dechema
Monograph, 1982): It is a stirred tank in which all dimensions are standardized
(cf. Table 3.7).
For modeling purposes, on the other hand, the reactor used should exhibit
pseudohomogeneity (see Sect. 4.3); such a reactor is called "perfect bioreactor"
(A. Moser, 1977b, 1983b). Several constructions are used. Atkinson (1974)
pointed out that only the biological film fermenter (BFF) configurations are
able to separate biological from physical parameters. Normally, in conven-
tional stirred tank fermenters operating with flocs, external and internal
transport limitations cannot be completely excluded.
Figure 3.42 shows several bioreactors suitable as perfect bioreactors for
quantifying biological rates unaffected by physical transport: (a) the sloping
plane biological film fermenter (BFF) of Atkinson (1974), (b) the thin-layer
tubular biofilm fermenter (ThLTBFF) (Moser, 1977a), (c) the horizontal
rotary BFF (Hoehn and Ray, 1973; Tomlinson and Snaddon, 1966) similar
to the horizontal rotary fermenter of Phillips et aI. (1961), (d) the vertically
rotating drum of Kornegay and Andrews (1968) and subsequently used by
La Motta (1976); and (e) the completely mixed microbial film fermenter
(CMMFF) of Atkinson and Davies (1972). The so-called multipurpose bio-
reactor (MBR, 1982; Moser, 1983b) includes several concepts of floc and film
bioprocessing: Both vertical and horizontal positions are possible using dif-
ferent stirring devices (impellers, rotating drums of different shape) Fig. 3.42f.
Another type of floc bioreactor is the completely filled bioreactor, which
realizes pseudo homogeneity to a high degree (Karrer, 1978; Puhar et aI., 1978).
Biofilm thickness and density are important factors for quantification. Bio-
logical film thickness can be measured using an optical procedure (Kornegay
and Andrews, 1968; Sanders, 1966; Trulear, 1980) or by means of electrical
conductance (Hoehn and Ray, 1973; Norrman et aI., 1977; see also Charaklis
et aI., 1982). The biofilm thickness can be controlled by mechanical means
(Atkinson, 1974) or hydrodynamically in the CMMFF or in the ThLTBFF.
The special property of a perfect reactors is illustrated in the case of GIL
processing, where the OTR and kLl . a are to be included in modeling. Both
terms, kLl and a, are affected by biological reactions in different ways, so that
GIL reactors with constant value of a are to be preferred. Figure 3.43 shows
three known configurations of GIL-model reactors with a = constant and
exhibiting no (or a negligible) velocity profile at the GIL interface (penetration
depth). This profile is of interest, because mass transfer, according to Equ. 2.3c,
is affected by fluid velocity, and the simple mass transfer theories mentioned
in Equ. 3.30b and c are only valid in the case of constant flow velocity with no
or a known gradient. As indicated in Fig. 3.43a, the laminar film on a vertical
I.;J
TABLE 3.7. Recommendations for bioreactor dimensioning.

'~"
Bench scale: Pilot scale: Technical scale:
ii-
Sl
Variable Recommendation* VR = 421 VR = 3001 VR = 30001 Q:j
o
Material r no. 4541,4571 Same Same
Pressure [bar] 4 Same Same
Temperature [DC] 143 Same Same
Height/diameter s 3: 1 3: 1 (2: 1) 3:1 (2:1)
S
dT [mm] s 259 492 1072
I
HT[mm] s 769 1595 3390
Heating Jacket Double jacket Double jacket or coils [
Motor drive Single, from bottom Same Same 1\0
Power [kW] 1.5 5 20 8-

'"d
Shaft-bearing assemblies r Double-crane pack Same Same t:::
Impeller (turbine): Changeable S
Vl
Diameter (impeller) di di = 0.4dT [
Width 0.3 'dT C'Il
Height 0.2dT 0'

....
Diameter (plate) 0.6dT Same Same '"d
No plates 6 a
No impellers 3 ~
Distance from bottom 0.3 dR,i
Distance to air distributor [mm] 15 40 60 ~
Rotational speed [rpm] (thyristorized motor) 0-2500 0-1200 0-500 g.
Air distributor (d = 0.7' d i ) s 3 jets each at 3 mm ddist = 25 mm with 10 holes ddisl =
50 mm with 100 holes
eachat4mm each at 4 mm
Maximum air flow rate [Nm 3 'm- 3 'min- I ] r 4 2 2
* r, recommended; s, strictly recommended. From Dechema, 1982.

I
tv
-J
-
128 3. Bioreactors
l--"
@ ~ ij- b
""'i8';" 'H"i" ':""~l

~~:,,:",:::-:':;'~:- ;-;:~;'t.:-;~i.!~:,.~\;'. ~!;":{.!':~? -:;.l:;~!i("..o;~~~
c
( ) L
- .1
I d
l ~
rl)
~: e
3.6 "Perfect Bioreactors" in Bench and Pilot Scale for Process Kinetic Analyses 129
... L
!
G L
FIGURE 3.43. GI L model reactors for systematic process kinetic analyses. (a) Falling
liquid film with thickness bL . (b) Falling liquid jet. (c) Thin-layer reactor with b =
penetration depth of gas, being always higher than bL (see A. Moser, 1973c). In all three
cases the velocity profile of liquid flow is simple enough to allow the application of
simple mass transfer theories.
<J FIGURE 3.42. Biofilm-model reactors and so-called perfect reactors for a systematic
process kinetic analysis. (a) Sloping plane biofilm reactor, according to Atkinson
(1974). (b) Thin-layer fermenter, according to Gorbach (see Moser, 1977a). (c) Rotating
cylinder biofilm reactor, according to Tomlinson and Snaddon (1966). (d) Rotating
drum, according to Kornegay and Andrews (1968). (e) Completely mixed microbial
film fermenter, according to Atkinson (1974). (f) Multipurpose bioreactor, according
to Moser (1983b), including case b (fJl and a horizontal stirred tank (f2) (Bauer and
Moser, 1985).
130 3. Bioreactors
plate exhibits the well-known half-parabolic profile (Beek and Muttzal, 1975),
while the laminar falling liquid jet (Fig. 3.43b) has a nearly uniform profile
(Danckwerts, 1970). The situation in the horizontal rotating drum in the
thin-layer tubular film fermenter (ThLTFF) is shown in Fig. 3.43c, where
the velocity profile in the film can-by calculation-be neglected in the
penetration depth of oxygen, b02 (Moser, 1973c).
The last type of ThLTFF is the most suitable one for aseptic bioprocessing.
It offers favorable properties for systematic process kinetic analyses, because
the GIL interface is given by geometry and kLl follows strictly a theory of
mass transfer (Equ. 3.30c), so that both terms need not be experimentally
measured. The significant role of "model bioreactors" for physiological studies
is recently stressed (Moser, 1988).
BIBLIOGRAPHY
Aiba, S., Humphrey, A.E., and Millis, N. (1976). Biochemical Engineering. New York:
Academic Press. 2nd edition.
Aiba, S., Nagai, S., and Nishizawa, Y. (1976). Biotechnol. Bioeng., 18, 1001.
Aiba S., Huang S.Y. (1969) Chern. Engng. Sci., 24, 1149.
Andre, G., et al. (1981). Biotechnol. Bioeng., 23,1611.
Andrews, G.F. (1982). Biotechnol. Bioeng., 24, 2013.
Atkinson, B. (1980). Presented at Conference on Biological Fluidized Bed Treatment
of Water and Wastewater. University of Manchester, England, 14-17 April.
Atkinson, B. (1974). Biochemical Reactors. London: Pion.
Atkinson, B., and Davies, 1.1. (1972). Trans. Inst. Chem. Engrs., 50, 208.
Atkinson, B., and Knight, A.J. (1975). Biotechnol. Bioeng., 17, 1245.
Atkinson, B., and Kossen, N.W.F. (1978). In Proceedings of the 1st European Congress
on Biotechnology, Interlaken, Switzerland, Dechema Monograph., 82, 37.
Atkinson, B., and Ur-Rahman, F. (1979). Biotechnol. Bioeng., 21, 221.
Baader, W., et al. (1978). Biogas in Theorie und Praxis. Hrsg. Kuratorium fUr Technik
und Bauwesen in der Landwirtschaft, Miinster-Hiltrup, West Germany.
Babivin, L.A., et al. (1981). Eur. J. Appl. Microbiol. Biotechnol., 13, 15.
Bailey, J.E. (1973). Chem. Eng. Commun., 1, 111.
Bajpaj, R.K., and Reuss, M. (1982). Canad. J. Chem. Eng., 60, 384.
Baker, C.G.J., et al. (1980). In Moo-Young, M., et al. (eds.). Advances in Biotechnology,
Vol. 1. Oxford: Pergamon Press, p. 635.
Bandyopadhyay, B., Humphrey, A.E., and Taguchi H. (1967). Biotechnol. Bioeng.,
9,533.
Bauer, A., and Moser, A. (1985). In Proceedings of the 5th European Mixing Conference
in Wiirzburg: BHRA Press Cranfield, Bedford, England, (J. Stanbury, ed.) p. 171
Bayer, K., and Fuehrer, F. (1982). Proc. Biochem., July/August, 42.
Beek, W.J. (1969). "Stofftransport mit und ohne chemische Reaktion," VSSD Skriptum.
Technical University, Delft, Netherlands.
Beek, W.J., and Muttzall, K.M.K. (1975). Transport Phenomena. New York and London:
John Wiley.
Beyeler, W., et al. (1981). Eur. J. Appl. Microbiol. Biotechnol., 13, 10.
Beyeler, W., et al. (1983). Chem. Ing. Techn., 55, 869.
Biggs R.D. (1963). AIChE Journal 9, 636.
Blanch, H.W. (1979). Annu. Rep. Ferrn. Proc., 3,47.

Bienke, H. (1979). Adv. Biochern. Eng., 13, 121.
Bronn, W.K. (1971). Chern. Ing. Techn., 43, 70.
Brown, D.E. (1967). Proc. Biochern., 2, 27.
Brown, D.E. (1977). Chern. Ind., 20 (August), 684.
Brown, D.E. (1984). In Process Variables in Biotechnology. Bioreactor Performance
working party (chairman W. Crueger). Frankfurt: Dechema, Chap. 17, p. 94.
Brown, D.E., et a!. (1979). Biotechnol. Lett., 1, 159.
Bruxelmane M. (1983). Technique de l'Ingenieur 2, A5910.
Bryant, J. (1977). Adv. Biochern. Eng., 5, 101.
Bryant, J., and Sadeghzadeh, N. (1979). Paper F3 presented at 3rd European Conference
on Mixing, York, England.
Calderbank, P.H. (1958). Trans. Inst. Chern. Eng., 36,443.
Carberry, J. (1964). Ind. Eng. Chern., 56, 39.
Cardoso-Duarte, J.M., et a!. (1975). In (Dean, A.c.R., et a!. (eds.). Continuous Culture,
Vol. 6. London: SCI Chap. 3.
CharakIis, W.G., et al. (1982). Water Res., 0,1.
CharakIis, W.G., et al. (1981). Biotechnol. Bioeng., 23, 1923.
Chibata, T., et al. (1972). In Terui, G. (ed.). Fermentation Technology Today. Tokyo:
Soc. of Fermentation Techno!., p. 383.
Cleland, N., et al. (1984). Appl. Microbiol. Biotechnol., 20, 268.
CMEA Recommendation (1968). Standard Methods of Water Analysis, Vol. 1. Budapest:
Comecon Vituki (Scientific Research Institute for Water Economy) Vo!' 1.
Cooker B. et a!. (1983). The Chern. Eng. 392, 81.
Cooney, Ch.L., Wang, D.I.C., and Mateles, R.I. (1968). Biotechnol. Bioeng., 11,269.
Cooper, C.M., Fernstrom, G.A., and MiIIer, S.A. (1944). Ind. Eng. Chern., 36, 504.
Coughlin, R.W., Charles, M., Paruchuri, E.K., and Allen, B.R. (1975). Chern. Ing. Techn.,
47, 111.
Cow, J.S., Littiehailes, J.D., Smith, S.R.L., and Walter, R.B. (1975). In Tannenbaum,
S.R., and Wang, D.I.C. (eds.). Single Cell Protein, II. Cambridge, Mass.: MIT Press,
p.424.
Danckwerts, P.V. (1958). Chern. Eng. Sci., 8, 93.
Danckwerts, P.V. (1970). Gas-Liquid Reactions. New York: McGraw-Hill.
Dang, N.D.P., Karrer, D.A., and Dunn, I.J. (1977). Biotechnol. Bioeng., 19,853.
Dawson, P.S.S. (1974). In Sikyta, B., Prokop, A., and Novak, M. (eds.). Advances in
Microbial Engineering, Part 2. New York Interscience Publications, p. 809.
Dechema (1982). Arbeitsrnethoden fur die Biotechnologie. Biotechnology working
group (chairman H.J. Rehm). Frankfurt: Dechema.
Dechema (1984). Process Variables in Biotechnology. Bioreactor Performance, working
party EFB (chairman W. Crueger). Frankfurt: Dechema.
Dechema (1987). Physical Aspects of Bioreactor Performance. Bioreactor Performance,
working party EFB (W. Crueger et aI., eds.). Frankfurt: Dechema.
Deckwer, W.D. (1977). Chern. Ing. Techn., 49, 213.
Deckwer, W.D. (1979). Fort. Verfahrenstechnik, 17(D}, 317.
Dohan, L.A., and Weinstein, H. (1973). Ind. Eng. Chern. Fundarn., 12,64.
Dunn, 1.1., and Mor, J.R. (1975). Biotechnol. Bioeng., 17, 1805.
Dunn, 1.1., and Einsele, A. (1975). J. Appl. Chern. Biotechnol., 25, 707.
Driessen, F.M., Ubbels, J., and Stadhouders, J. (1977). Biotechnol. Bioeng., 19,841.
Diidiikovic, M.P. (1977). Ind. Eng. Chern. Fundarn., 16,3,385.
132 3. Bioreactors
Ebert, KH. (1971). Chem. Ing. Tech., 43,50.

Einsele, A. (1976a). Ph.D. Thesis, Technical University, Zurich.
Einsele, A. (1976b). Chem. Rundschau, 29, 53.
Einsele, A., and Fiechter, A. (1974). Chem. Ing. Tech., 46,701.
Einsele, A., Ristroph, D.L., and Humphrey, A.E. (1978). Biotechno!. Bioeng., 20,1487.
Einsele A. and Finn R.K (1980). Ind. Engng. Chem. Proc. Des. Dev., 19,600.
Eriksson, R. (1971). In Proceedings of the 1st European Biophysics Congress, Medical
Academy. Vienna: Verlag der Wiener medizinischen Akademie (Broda E. et aI., eds.)
IV, 319.
Eriksson, R., and Holme, T. (1973). Biotechno!. Bioeng. Symp., 4, 581.
Falch, E. (1968). Biotechnol. Bioeng., 10,233.
Fan, L.T., Erickson, L.E., Shah P.S., and Tsai, B.I. (1970). Biotechnol. Bioeng., 12, 1019.
Fiechter, A. (1974). In Proceedings of the 4th International Symposium on Yeasts,
Part II. Vienna: University Bodenkultur Press (Klaushofer H., Sleytr U. eds.), 17.
Fiechter, A. (1978). In Proceedings of the 1st European Congress on Biotechnology
Interlaken, Switzerland. Dechema Monograph, 82,17.
Finn, R.K (1969). Proc. Biochem., 4, 17.
Ford, J.R., et ai. (1972). Biotechnol. Bioeng. Symp., 3, 267.
Forrest, W.W. (1972). In Norris,J.R., and Ribbons, D.W. (eds.). Methods in Microbiology,
Vol. 6B. London-New York: Academic Press, p. 285.
Fu, B., Weinstein, H., Bernstein, B., and Shaffer, A.B. (1971). Ind. Eng. Chem. Proc. Des.
Dev., 10,501.
Ghose, T.K, and Mukhopadhyay S.N. (1979). Paper presented at 32nd Annual Session
ofInternational Institutes of Chemical Engineering, Indian Institute of Technology-
lIT, New Delhi, in Bombay.
Grieves, R.B., et al. (1964). J. App!. Chem., 14,478.
Griot, M., et al. (1986). In Proceedings Intern. Conf. on Bioreactor Fluid DynamiCS.
(Cambridge, April): BHRA Press, Granfield, Bedford, England (J. Stanbury., ed.)
p.203
Grosse, H.H., et al. (1985). Acta Biotechnolog., 5, 163, (esp. ref. 5).
Hartung, KH., and Hiby, J.W. (1973). Chem. Ing. Techn., 45, 522.
Heineken, F.G. (1970). Biotechnol. Bioeng., 12, 145.
Heinzle, E. (1978). Ph.D. Thesis, Technical University, Graz, Austria.
Herbert, D., et al. (1956). J. Gen. Microbiol., 14,601.
Herzog, P., et ai. (1981). Chem.Ing. Techn., 55, 566.
Hesseltine, e.W. (1977a). Proc. Biochem., 11 (July-August), 24.
Hesseltine, e.W. (1977b). Proc. Biochem., 11 (November), 29.
Hiby, J.W. (1972). Chem. Ing. Techn., 44, 907.
Hines, D.A., Bailey, M., Onsby, J.e., and Roesler, F.e. (1975). 1st Chern. Eng. Symp.,
Ser. 41, Dl.
Hirner, W. (1974). Ph.D. Thesis, Technical University, Stuttgart.
Hirner, W., and Blenke, H. (1974). Chem. Ing. Techn., 46, 353.
Hoehn, R.e., and Ray, A.D. (1973). J. Water Poll. Contr. Fed., 45, 2302.
Holmes, D.B. (1964). Chern. Eng. Sci., 19,201.
Howell, J.A., and Atkinson, B. (1976). Biotechno!. Bioeng., 18, 15.
Imanaka, T., and Aiba, S. (1976). J. Appl. Chem. Biotechno!., 26, 559.
Jones, P.H. (1973). Proc. Biochem., September, 19.
Joshi J.B. (1980). Trans. Inst. Chem. Engrs., 58, 155.
Joshi J.B., Pandit A.B., and Sharma M.M. (1982). Chem. Eng. Sci., 37, 813
Jiittner, F. (1982). Process Biochem., March/April, 2.

Kiippel, M. (1976). VDI Bericht, 578.
Kiippeli, 0., and Fiechter, A. (1980). Biotechnol. Bioeng., 22,1509.
Kiippeli, 0., and Fiechter, A. (l981a). Biotechnol. Bioeng., 23,1897.
Kiippeli, 0., and Fiechter, A. (1981 b). Biotechnol. Lett., 3, 541.
Karrer, D. (1978). Ph. D. Thesis, Technical University, Ziirich.
Katinger, H.W.D. (1976). Mitteil. d. Versuchsanstalt fUr das Giirungsgewerbe, Wien
Nr. 7/8, 82
Katinger, H.W.D., Scheirer, W., and Kromer, E. (1979). Ger. Chem. Eng., 2, 31.
Khang, S.J., and Levenspiel, o. (1979). Chem. Eng. Sci., 31, 569.
King, c.J. (1966). Ind. Eng. Chem. Fundam., 5, 1.
Kipke, K.D. (1984). In Dechema Monograph (1984). Process Variables in Biotechnology.
Bioreactor Performance EFB-working party (W. Crueger chairman). Frankfurt:
Dechema, Chap. 20.
Kishinevskii, M.Kh. (1951). J. Appl. Chem. (USSR), 24, 593.
Kono, H. (1980). Hydrocarb. Proc., January, 123.
Korjagin, W.W., et al. (1978). In Entwicklung von Laborfermentoren. 6th Reinhards-
brunner Symposium, Academy of Science. Berlin: Akademie Verlag, Ringpfeil M.,
ed. p. 327.
Kornegay, B.H., and Andrews, J.F. (1968). J. Water Poll. Contr. Fed., 40, 460.
Kramers, H., and Alberda, G. (1953). Chem. Eng. Sci., 2, 1731.
Kramers, H., and Alberda, B. (1953). Chem. Eng. Sci., 2, 35
Kiing, W., and Moser, A. (1986). Bioproc. Eng., 1,23.
Kiing, W., and Moser, A. (1988). Biotechnol. Lett., in press.
Liiderach, H., Widmer, F., and Einsele, A. (1978). Proceedings of the 1st European
Congress on Biotechnology, Part I, Interlaken, Switzerland. Dechema Monograph, 82.
La Motta, E.J. (1976). Biotechnol. Bioeng., 18, 1359.
Lee, Y.H., and Tsao, G.T. (1979). Adv. Biochem. Eng., 13,35.
Lehmann, J., et al. (1980). In Moo-Young, M., et al. (eds.). Proceedings 6th International
Fermentation Symposium (Advances in Biotechnology), Vol. 1. Oxford: Pergamon
Press, p. 453.
Lehmann, J., et al. (1982). In Arbeitsmethoden fur die Biotechnologie. Working group
on Biotechnology (H.J. Rehm, chairman). Frankfurt: Dechema.
Lehmann, J., et al. (1985). In Rehm, H.J., and Reed, G. (eds.). Biotechnology. Vol. 2.
Deerfield Beach, Fla., and Basel: Verlag Chemie Weinheim, Chap. 25.
Lehnert, J. (1972). Velfahrenstechnik, 6, 58.
Leistner, G., Miiller, G., Sell, G., and Bauer, A. (1979). Chem. Ing. Techn., 51, 288.
Levenspiel, O. (1972). Chemical Reaction Engineering. New York: John Wiley.
Levenspiel, O. (1979). The Chemical Reactor Omnibook. Corvallis, Ore.: OSU Book
Stores.
Levenspiel, 0., and Smith, W.K. (1957). Chem. Eng. Sci., 6, 227.
Liepe, et al. (1978). In Proceedings of the 1st European Congress on Biotechnology,
Interlaken, Switzerland, Part 1, p. 78, Dechema Monograph, 82.
Linek, V., Sobotka, M., and Prokop, A. (1973). Biotech. Bioeng., 15,429.
Liu, M.S., et al. (1973). Biotechnol. Bioeng., 15,213.
Liicke, J., Oels, u., and Schiigerl, K. (1977). Chem. Ing. Techn.,49, 161.
Luft, G., and Herbertz, H.A. (1969). Chem. Ing. Techn., 41, 667.
Luong, J.H.T., and Volesky, B. (1980). Canad. J. Chem. Eng., 58,497.
Luong, J.H.T., and Volesky, B. (1982). Canad. J. Chem. Eng., 60,163.
134 3. Bioreactors
Marison, I.W., and von Stockar, U. (1983). In Proceedings BIOTECH 83, International
Conference on Commercial Applications & Implications of Biotechnology, Online
Pub!. Ltd., Northwood/UK, p. 947.
Marison, I.W., and von Stockar, U. (1987). Enzyme and Microbial Technology, 9, 33.
Markl, H., and Vortmeyer, D. (1973). In Proceedings of 3rd Symposium on Technical
Microbiology, Berlin: Dellweg H., ed., Garungsgewerbe und Biotechnologie p. 29.
Mathur, K.B., and Epstein, N. (1974). Spouted Beds. New York: Academic Press.
MBR (Microbial Bioreactor AG, in Wetzikon/Ziirich CH) (1982). In Neuentwicklungen,
ACHEMA 82 (Dechema, pub!.) 1,430.
Mersmann, A. et a!. (1976). Intern. Chern. Engng., 16,590.
Merz, A., Vogg, H. (1978). Chern. Ing. Techn. 50, 108.
Meyer, e., and Beyeler, W. (1984). Biotechnol. Bioeng., 26, 916.
Meyer H.P. (1987) in Physical aspects of Bioreactor Performance, EFB-working party
Bioreactor Performance, (Crueger W. et a!., eds.) Dechema/Frankfurt D., Chap. 8.
Meyrath, J., and Bayer, K. (1973). In Proceedings of the 3rd Symposium on Technical
Microbiology. Berlin: Dellweg, H. (ed.), Inst. fUr Garungsgewerbe und Biotech-
nQlogie. p. 117.
Middleton, J.e. (1979). Paper A2 presented at 3rd Duropean Conference on Mixing,
York, England.
Midler, M., and Finn, R.K. (1966). Biotechnol. Bioeng., 8, 71.
MihaItz, P., and Hollo, J. (1980). Acta Biotechnolog., 0, 39.
Minkevich, I.G., and Eroshin, V.R. (1973). Folia Microbiol., 18,376.
Moes, J., et a!. (1984). In Proceedings of the 3rd European Congress on Biotechnology.
Munich: Vo!' 2, p. 285. Verlag Chemie, Weinheim/D.
Monk, P.R. (1978). Proc. Biochem., December, 4.
Moo-Young, M. et a!. (1979). Biotechnol. Bioeng., 21,593.
Moo-Young, M. (1985). (ed.-in-chief). Comprehensive Biotechnology, Vo!. 1-3. Oxford:
Pergamon Press.
Moser, A. (1973a). Biotechnol. Bioeng. Symp., 4, 399.
Moser, A. (1973b). In Proceedings of the 3rd Symposium on Technical Microbiology.
Berlin: Dellweg, H. (ed.) Inst. fUr Garungsgewerbe u. Biotechnologie. p. 62.
Moser, A. (1973c). Chem. Ing. Techn.,45, 1313.
Moser, A. (1974). Gas Wasserfach/Wasser-Abwasser, 115,411.
Moser, A. (1977). Chem. Ing. Techn., 49,612.
Moser, A. (1978). In preprints, 1st European Congress on Biotechnology, Interlaken,
Switzerland, Part 1, 88.
Moser, A. (1980a). In Ghose, T.K. (ed.). Proceedings of Bioconversion and Biochemical
Engineering Symposium, Vo!. 2, New Delhi: Indian Institute of Technology, New
Delhi p. 253.
Moser, A. (1980b). In Theoretical Basis of Kinetics of Growth, Metabolism and Product
Formation of Microorganisms, Part 2. UNEP/UNESCOjICRO Training Course,
Zentralinstitut fUr Mikrobiologie und Experimentelle Therapie, Academy of Science,
Jena, (Knorre W., ed.) East Germany (publisher) p. 27.
Moser, A. (1983a). Biotechnol. Lett., 4, 73.
Moser, A. (1983b). Proceeding Adv. in Ferm., 83 (Supp!. to Process Biochemistry) p. 202.
Moser, A. (1985a). In Moo-Young, M. (ed.-in-chief). Comprehensive Biotechnology,
Vo!. 2. Oxford: Pergamon Press, Chap. 4.
Moser, A. (1985b). In Proc. Adv. Bioreactor Engng., L0dz, Poland, September 1985.
Moser, A. (1987). In Physical Aspects oj Bioreactor PerJormance. EFB-working party
Bioreactor Performance (Crueger W., et aI., eds.). Dechema. Frankfurt; Chap. 4.
Moser, A. (1988). Trends in Biotechnology, Vol. 6, No. 11.
Moser, A., and Steiner, W. (1974). Chem. Ing. Techn., 46,695.
Moser, A., and Steiner, W. (1975a). Chem. Ing. Techn., 47, 211.
Moser, A., and Steiner, W. (1975b). VDI Bericht, 232, 259
Moser, A., Kosaric N., and Margaritis A. (1980). Paper presented at 30th Canad. Chern.
Eng. Conference, Edmonton, Canada, October.
Moser, A., Kung, W., Weiland, P. (1985). Paper presented at Achema/Frankfurt. In
Dechema preprints Biotechnology ACHEMA, Intern. Meeting of Chern. Engng.
Moser, F. (1977). VerJahrenstechnik, 11, 670.
Mou, D.G., and Cooney, Ch.L. (1976). Biotechnol. Bioeng., 18, 1371.
Mukataka, S. et al. (1980). J. Ferm. Technol, 58, 155.
Mukataka, S. et aI. (1981). J. Ferm. Techno!., 59, 303.
Mulcahy, L.T., and La Motta, E.J. (1978). Rep. No. Env. E. 59-78-2, Dept. of Civil
Engng., University of Massachusetts, Amherst.
Nagel, 0., and Hegner, B. (1973). Chem. Ing. Techn., 45, 913.
Nagel, 0., Kurten, H., and Sinn, R. (1972). Chem. Ing. Techn., 44, 367.
Nelbock, M., and Wandrey, C. (1978). In: Biotechnologie. Frankfurt: Umschau-Verlag.
Bundesministerium f. Forschung & Technologie, Bonn, p. 81.
Ng, D.Y.C., and Rippin, D.W.T. (1964). Proceedings oj the Third European Symposium
on Chemical Reaction Engineering Amsterdam. Oxford: Pergamon Press, p. 161.
(van Heerden C. et aI., eds.)
Nikolaev, P.I., et aI. (1976). Theoret. Found. Chem. Eng. (TransI.), 10, 13.
Norrman, G., et al. (1977). Develop. Ind. Microbiol., 18,581.
Norwood, K.W., and Metzner, A.B. (1960). AlChE Journal, 6, 432.
Ohyama, Y., and Endho, K. (1955). Chem. Eng. (Jap.), 19,2.
Oosterhuis, N.M.G. (1984). Ph.D. Thesis, Technical University, Delft, Netherlands.
Paar H., et al. (1988). Bioprocess Eng., 3, in press.
Parker, D.S., Kaufmann, W.J., and Jenkins, D. (1971). J. Water Poll. Contr. Fed.,
43,1817.
Paspek, St. G., et al. (1980). Chem. Eng. Educ., Spring, 78.
Pawlowski, J. (1962). Chem. Ing. Techn., 34, 628.
Philipps, K.L. (1969). In Perlman, D. (ed.). Fermentation Advances. New York: Academic
Press, p. 465.
Philipps, K.L., et aI. (1961). Ind. Eng. Chem., 53, 749.
Pickett, A.M., Topiwala, H.H., and Bazin, M.J. (1979). Proc. Biochem., 13 (Nov., 10).
Pirt, S.J. (1974). J. App!. Chem. Biotechnol., 24,415.
Pirt, S.J. (1980). Plenary Lecture, 6th Internation Fermentation Symposium London,
Ontario. (Zajic J.E. et aI., eds.) PubI. Sales & Distribution, National Research
Council Canada (Ottawa).
Pirt, S.J., et aI. (1983). J. Chem. Techn. Biotechnol., 33B, 35.
Pitcher, W.H., Jr. (1978). Adv. Biochem. Eng., 10, 1.
Popovic, M., Niebelschutz, H., and Reuss, M. (1979). Eur. J. Appl. Microbiol. Biotechnol.,
8, 1.
Powell, 0., and Lowe, J.R. (1964). In Malek, I., (ed.). Continuous Culture oj Micro-
organisms. Prague: Czechoslovakia Academy of Science, p. 45.
136 3. Bioreactors
Prochazka J. and Landau J. (1961) ColI. Czech. Chern. Commun. 26, 2961.
Puhar, E., Karrer, D., Einsele, A., and Fiechter, A. (1978). In 1st European Congress on
Biotechnology, Interlaken, Switzerland, Part 2, 1/83. Dechema, Frankfurt.
Pye, E.K., and Wingard, L.B., Jr. (1974ff). Enzyme Engineering, Review Series. New
York: Plenum Press.
Quicker, G., et al. (1981). Biotechnol. Bioeng., 23, 635.
Rehm, H.I. (1980). Technische M ikrobiologie. Heidelberg-Berlin-New York: Springer-
Verlag.
Rehm, H.J., and Reed, G. (eds.) (1982ff). Biotechnology-A Comprehensive Treatise,
Vol. 1-9, esp. Vol. 2. Deerfield Beach, Fla., and Basel: Verlag Chemie Weinheim.
Reith, T. (1968). Ph.D. Dissertation, Technical University, Delft, Netherlands.
Reuss, M. (1976). In Proceedings of the 5th International Fermentation Symposium,
Berlin. Dellweg H. (ed.), Inst. f. Garungsgewerbe und Biotechnologie, Berlin.
Reuss, M. (1977). In Proceedings of the Tutzing Symposium. Dechema Monograph, 81,
45. Rehm H.J. (ed.) Verlag Chemie, Weinheim.
Reuss, M., and Bajpaj, R.K. (1988). In Ho, Ch., and Wang, D.I.e. (eds.). In Biochemical
Engineering: Bioreactor Design and Operation. Butterworth. Stoneham MA., in
press.
Reuss, M., and Wagner, F. (1972). In Dechema Monograph, 71, 9.
Richards, J.W. (1968). Introduction to Industrial Sterilization. London-New York:
Academic Press.
Rippin, D. (1967). Ind. Eng. Chem. Fundam., 6, 488.
Ruchti, G., et al. (1981). Biotechno!' Bioeng., 23, 277.
Rudkin, 1. (1967). Br. Chem. Eng., 12, 1374.
Rushton, J.H., Costich, E.W., and Everett, H.J. (1950). Chem. Eng. Progr., 46, 467.
Ryu, D., and Humphrey, A.E. (1972). J. Ferm. Techno!., 50, 424.
Sanders, W.M. (1966). AirWater Poll. Int. J., 10,253.
Sano, Y., Yamaguchi N., and Adachi, T. (1974). J. Chern. Eng. (Japan), 7, no. 4, 255
Schauerte, W.A. (1981). Ph.D. Thesis, Technical University, Munich.
Schmidt, K.G., and Blenke, H. (1983). Verfahrenstechnik, 17,593.
Schneider G., Purgstaller A., Somitsch H. and Moser A. (1986). in Proc. Biochemical
Engrg. Congress Stuttgart, (Chmiel H. et aI., eds.) G. Fischer Verlag, p. 928.
Schneider H., and Moser, A. (1984). Biotechnol. Lett., 6, 295.
Schneider, H. and Moser, A. (1985), ibid 7, 376.
Schreier, K. (1975). Chemiker Zeit., 99, 328.
Schugerl, K. (1977). Chem. Ing. Techn., 49,605.
Schugerl, K. (1979). In Jottrand, R. (ed.). Ausbildungskurs uber "Chemische Verfahren-
stechnik in den biologischen Operationen" Brussel, November. University Press
Brussels.
Schugerl, K. (1980). Chem. Ing. Techn., 52, 951.
Schugerl, K. (1982). Adv. Biochem. Eng., 22, 94.
Schugerl, K., Lucke, J., Lehmann, J., and Wagner, F. (1978). Adv. Biochem. Eng., 8, 63.
Schugerl, K., Oels, U., and Lucke, J. (1977). Adv. Biochem. Eng., 7, 1.
Schumpe, A., et al. (1982). Adv. Biochem. Eng., 24, 2.
Schumpe, A., et al. (1985). In Rehm, H.1., and Reed, G. (eds.). Biotechnology, Vol. 2,
Deerfield Beach, Fla., and Basel: Verlag Chemie Weinheim, Chap. 10.
Sinclair, C.G. (1984). Biotechnol. Lett., 6, 65.
Sinclair, e.G., and Brown, D.E. (1970). Biotechnol. Bioeng., 12, 1001.
Sittig, W. (1977). Fort. Verfahrenstechnik, 15D, 354.
Sittig, W., and Heine,H. (1977). Chem. Ing. Techn.,49, 595.

Sobotka, M., et al. (1982). In Tsao, G.T. (ed.). Annual Report on Fermentation Processes.
New York: Academic Press.
Steiner, W., et al. (1977). In 4th FEMS Symp. (Abstr. B33). Federation of European
Microbiological Societies, Bu'Lock, J., Meyrath, J. (eds.) University Bodenkultur
Press.
Stenberg, 0., (1984). Ph.D. Thesis, Technical University, Goteborg, Sweden. (Prof.
N.H. Schoon).
Taguchi, H., and Humphrey, A.E. (1966). J. Ferm. Technol., 44, 881.
Tanaka, H. et al. (1975a). J. Ferm. Technol.. 53, 18.
Tanaka, H. et al. (1975b). J. Ferm. Technol.. 53, 35.
Toda, K., and Dunn, 1.1. (1982). Biotechnol. Bioeng., 24, 651.
Tomlinson, T.G., and Snaddon, D.M. (1966). Air Water Poll., 10, 865.
Trulear, M.G. (1980). MSc. Thesis, Rice University, Houston.
Tsai, B.I., Erickson, L.E., andFan, L.T. (1969). Biotechnol. Bioeng., 11, 181.
Tsai, B.I., Fan, L.T., Erickson, L.E., and Chen, M.S.K. (1971). J. Appl. Chem. Biotechnol.,
21,307.
van de Vusse, J.G. (1955). Chem. Eng. Sci., 4, 178.
van Krevelen, D.W., and Hooftijzer, P.J. (1948). 21st Congress International de Chimie
Industrielle Bruxelles, numero speciale, Chimie et Industrie. p. 168.
van Meyenburg, K., and Fiechter, A. (1968). Biotechnol. Bioeng., 10, 535.
van Stockar, u., and Stravs, A.A. (1983). Swiss Biotech., 6, 7.
van Uden, N. (1971). Z. Allgem. Mikrobiol., 11,6,541.
Venkatsubramanian, K. (ed.) (1979). Immobilized Microbial Cells. ACS Symposium
Series 106. Washington, D.C.: American Chemical Society.
Volesky, B., et al. (1982). J. Chem. Tech. Biotechnol., 32, 650.
Votruba, J., and Sobotka, M. (1976). Biotechnol. Bioeng., 18, 1815.
Wandrey, Ch. (1983). In Proceeding "Biotech 83". International Conference on Com-
mercial Applications & Implications of Biotechnology, Online Publ. Ltd., North-
wood/UK, p. 577.
Wandrey, Ch., and Flaschel, E. (1979). Adv. Biochem. Eng., 12, 148.
Wang, D.I.C., Cooney, Ch.L., Demain, A.L., Dunnill, P., Humphrey, A.E., and Lilly,
M.D. (1979). Fermentation and Enzyme Technology. New York: John Wiley.
Wang, D.I.C., and Humphrey, A.E. (1969). Chem. Eng., 76, 108.
Wang, H., et al. (1978). Eur. J. Appl. Microbiol. Biotechnol., 5, 207.
Weinstein, H., and Adler, R.J. (1967). Chem. Eng. Sci., 22, 65.
Wen, C.Y., and Fan, L.T. (1975). Models for Flow Systems and Chemical Reactors.
New York: Marcel Dekker.
Werther, J. (1977). Chem. Ing. Techn., 49, 777.
Werther, J. (1980). Chem. Eng. Sci., 35, 372.
Zlokarnik, M. (1967). Chem. Ing. Techn., 39, 539.
Zlokarnik, M. (1975). Verfahrenstechnik, 9, 442.
Zlokarnik, M. (1978). Adv. Biochem. Eng., 8,133.
Zlokarnik, M. (1979). Adv. Biochem. Eng., 11, 157.
Ziegler, H., Meister, D., Dunn, 1.1., Blanch, H.W., and Russel, T.W.F. (1977). Biotechnol.
Bioeng., 19, 507.
Ziegler, H. et al. (1980). Biotechnol. Bioeng., 22, 1613.
Zwietering, Th.N. (1959). Chem. Eng. Sci., 11, 1.
CHAPTER 4
Process Kinetic Analysis
4.1 Kinetic Analysis in Different Types of Reactors

A schematic diagram for a homogeneous biological process was shown in Fig.
2.3. For comparison, Fig. 4.1 shows a typical situation in a heterogeneous
system; in it a rotating disk binds a microbiological film. In analogy to Fig.
2.3, all significant process variables in the three reaction phases (gas, liquid,
and solid) are shown along with the limiting transport steps, in the order 1 to
4.
Mass transport between phases and the relative extent of each phase are
important in deciding whether and if a homogeneous or heterogeneous model
can be applied. In dealing both with microbiological particles and films, the
variable characterizing the size has a distribution function, as discussed in
Sect. 3.3.9: From the distribution of actual particle diameters one obtains dp ,
the mean diameter. Two extreme cases, of interest in process engineering, can
be distinguished: Uncontrolled thickness growth will be found in all bioreactors
in which shear forces vary greatly from place to place (stirred vessels). Un-
controlled growth is also normal in biofilm reactors, especially in trickling
filters.
Controlled thickness growth of particles would be conceivable if a homo-
geneous shear force field could be established. Initiatives in this direction can
be found in the literature (EPA, 1977; Moser, 1983; Schreier, 1975). The control
of size can be better accomplished, however, in film-type bioreactors. This is
done (a) using a mechanical scraping device of a particular height, as in
the "biological film reactor" (Atkinson, 1974), (b) using mechanical methods
involving surface friction to displace the biological film from the surface of
the small particles in a fluidized bed reactor (completely mixed microbial film
fermenter, Atkinson and Davies, 1972), or (c) using a method of controlling
the hydrodynamic stress with the aid of peripherical speed of a rotating drum,
which is the support for the growth of a particular biological strain (thin film
fermenter, Moser, 1977b). The characteristics of these film reactors with
controlled thickness makes them especially well suited for obtaining kinetic
data for bioprocesses (Atkinson, 1974), as outlined in Sect. 3.6.
To illustrate the complexity of real fermentation, Fig. 4.2 shows a plot of
4.1 Kinetic Analysis in Different Types of Reactors 139
DISK
a
o
GAS LIQUID FIU~
----++'i!~p.+---~ .. SHAFT
LIQUID BULK
s.I
FIGURE 4.1. Heterogeneous biological processes utilizing a biological film (a biodisk

reactor). A group of disks is fastened to a horizontal shaft. Each disk supports the
growth of the desired culture. The disks are turned through the liquid and gas phases.
Mass transport is subject to the same limitations as given in Fig. 2.3 for pseudo-
homogeneous processes. With the biodisk, however, step 4 (limitation in the solid
phase) becomes significant. (From Moser, 1981.)
f (d/d r )
0.2
.3
.4
FIGURE 4.2. Simulations of oxygen consumption kinet- .5
.6
ics qO/qO.max at different impeller/tank diameter ratios
d)d T as a function of power consumption P/V, as a
typical example of macrokinetics in a 20 I vessel. (From .2~--~--~2~--~3
Reuss et aI., 1982.) p/V KW.M 3
the respiration rate, qo, versus the specific power input at varied impeller tank
diameter ratio for a stirred rank reactor (VR = 20 1) as observed for mycelial
growth (Steel and Maxon, 1966) and later on simulated by Reuss et al. (1982).
The appearance of these pseudokinetics is the consequence of interactions
between physics (transport in a bioreactor influenced by geometry) and biol-
140 4. Process Kinetic Analysis
~'NTERACT'ONS~
No significant changes Significant

No cit profile interactions
(pseudohomogeneous cit and clz profiles
a/pproxima~n) ./ \
" Macroscopic Microscopic

Noclz With level; level;
profile clz profile extracellular intracellular
I
CSTR CPFR
I / "Transport
Transport " I
Transient
limitation enhancement operation
techniques
(balanced,
frozen,
External Internal true dynamics)
(k L2 ,k L1 ) (k s )
FIGURE 4.3. Classification of different types of interactions between microbial meta-

bolism, respectively kinetics, and physical transport phenomena in the environment
(bioreactor).
!
z
G
FIGURE 4.4. Schematic illustration of typical experimental situations in real technical

scale bioprocessing: transient conditions due to environmental changes caused by
gradients in concentrations, temperature, shear, and pressure in the reactor as well as
in the inflowing liquid stream.
4.2 Regime Analysis-General Concept and Guidelines 141
ogy. Figure 4.3 shows different types of interactions possible in bioprocessing.

With the exception of conventional batch processes, we have thus far con-
sidered only bioreactors operating under steady-state conditions with no
external perturbation. However, most industrial bioprocesses operate under
non stationary conditions, as illustrated in Fig. 4.4. Fluctuations in the inflow-
ing stream (e.g., waste water pollution), gradients of concentrations of com-
ponents (X, S, 0, C, P, Hv), pH, shear, pressure, and so on over the length or
depth of reactors are obvious, resulting in changes in the cell's environment
and composition.
The strategy presented in the following section enables bioprocess engineers
to approach such complex situations.
4.2 Regime Analysis-General Concept and Guidelines

By analogy to an interesting approach to the depiction of complex systems in
thermodynamics (Prigogine and Defay, 1954), the concept of characteristic
times of internal processes can be extended to the treatment of bioengineering
systems (Bailey, 1973; Harder and Roels, 1982; Kossen, 1979; Moser, 1977b,
1981, 1982, 1984; Oosterhuis, 1984; Roels, 1982; 1983; Sweere et al. 1987). A
systematic approach to problems involving interactions between the environ-
ment and cells seems to be possible for case 1, transport effects, including
transport to and in measuring electrodes; and for case 2, transient effects of
reactions, in which enzyme regulation phenomena such as induction and
repression must be considered (cf. Fig. 4.9). A guideline for quantification and
modeling can be summarized as follows (Moser, 1982):
1. Describe significant phenomena (see Table 4.1): case 1, transports: RTD,
mixing, CTD, OTR, HvTR, and so on, and/or case 2, metabolic rates:
significant and compositional changes in cell content, for example, proteins,
DNA, RNA, enzymes, macromolecules such as PHB, as internally balanced
units together with macroscopic process variables.
2. Determine characteristic rate constants or characteristic times from experi-
ments: case 1: mean residence time (f), mixing time (trn), cycle time (to>, OTR
time (t oTR ), and/or response time of measuring devices, for example, O 2
electrode (t E ); case 2: at macroscopic level, characteristic reaction times t r i ,
which are indirectly proportional to the specific rates (j1., qs, qo, qp, qc, qH);
and at microscopic level, individual time responses for changes in DNA,
RNA, protein enzymes, and so on (see Table 4.1).
3. Compare rate constants or, for significant phenomena, characteristic times
of changes over the period of interest, and determine hierarchy of individual
response time or rate constant ("regime"). Table 4.2 lists possible cases and
results of regime analysis in bioreactor operation.
A conclusion from these considerations on mathematical modeling of
biokinetics is that in both interaction cases (transport and metabolic changes),
TABLE 4.1. Characteristic values of k and t used in regime analyses of bioprocesses.

Macroscopic process variable kcharact tcharaa Typical range [s]
Kinetics (k, resp. t,)
X Pmax t~ = I/Jl.max 1.10 5 -1.104
S qs.max tq, 1 104 -5' 104
0 qo.max tqo 0'1-20
p qp.max tqp
C qc.max tqC
Hv
cp'p'I'1T
qH".H tqH =
, (rHJ + (rH,hR
Cell compartments qint tqinl
(internal compounds)
Transports (kTR resp. tTR )
Micromixing time tm asf(m) 10-3 .102
Circulation time tc 1-30
Circulation time distribution t.,S2
Mean residence time fL = VdFL
OTR global kL1a tOTR = I/kLl a 5-10
OTR gas bubble t8TR = He/kLl a 3 '102 -6 .10 2
_ VL
gas residence time to = (1 - 60)- 10-20
Fa
cpp
Heat transfer kH,aH, tH,TR=~ 3 '10 2 -8 '10 2
Hv Hv
Electrode response kE tE = I/kE
Substrate addition tSA = V'S/Fs ' So 10 1 -10 2
Concept of effectivity factors 1'/, = f(k,/k TR )
Transport limitation
Transport enhancement I'/T, = f(kT,/k R )
a drastic reduction of model complexity can be achieved with the aid of the
rds, respectively qss, principle.
Figure 4.5 provides an overview of the range of characteristic times of
significant phenomena in bioprocessing, including physical transport and
biological reactions on the macroscopic and microscopic level and indicating
regions where the organism and environment interact (Roels, 1983). Interest-
ing papers on this subject can be found in the literature (Bailey, 1973; Bazine,
1982; Harder and Roels 1982; Kossen, 1979; Pickett et ai. 1979; Roels, 1982;
Romanovsky et aI., 1974; Sweere et aI., 1987).
In cases involving complex situations, simple unstructured models are often
insufficient. Both bioreactor models and biological reaction models, there-
fore, must be structured to some degree. Figure 4.6 is a graphical representa-
tion of a two-compartment model for oxygen transfer in a stirred tank reactor,
where the liquid phases are structured into mixed and segregated (bubble)
zones (connected by a pumping capacity), while the gas phase is assumed to
be well mixed. A quite simple model for OTR is shown in Fig. 4.7, where
the liquid phase is separated into a stirrer zone with high hydrodynamic
stress and a zone of bulk flow (Oosterhuis, 1984; see also Sect. 6.9.4). A
TABLE 4.2. Consequences of comparison of character-

istic values in applying regime analysis to bioprocessing.
Comparison Consequence
fL tm Completely mixed reactor (L phase)

tm < tr,i Pseudohomogeneous reactor model
tc < tr,i Pseudohomogeneous reactor model
tOTR < tqO No O 2 limitation
tOTR ~ tqo O 2 limitation (02 gradients)
tg TR < fG Gas bubbles not depleted
tH,TR - t qH, Isotherm process
t. < tqH, No T gradient
tE < tOTR No electrode effect on OTR
'I, = 1 (k, < ~R) Kinetic regime
k, > ku External transport limitation (GIL)
k, > kL2 External transport limitation (LIS)
k, > k, Internal transport limitation (S phase)
'IT, < 1 (k, > ~R) Diffusion transport regime
'In> 1 Transport enhancement
t. t, "Balance" system (see Sect. 5.7)
t. t, "frozen" biosystem (see Sect. 5.7)
te "" tr Resonance, thus true dynamics
tq, - t S A S-limited growth
tm - t S A S gradients in reactor
tm - tOTR O 2 gradients in reactor
MASS ACTION
.. ENZYME CONTROL
ALLOSTERIE .. GENETI C CONTROL

ENZYME INDUCTION
.. SELECTION
.. EVOLUTION
10- 4
I
10- 2 100 102 104 'Tr
"Te
.. (S)
MIXING GRADIENTS
DYNAMICS (DCSTR,SCSTR,
CSTR-TRANSIENTS)
STEADY STATES

CONT,CULTURES
FIGURE 4.5. Comparison of the relaxation times (characteristic response times, see
Table 4.2), of the reaction mechanisms inside organisms ('t,) and those of characteristic
times of environment in bioreactors ('t e ), according to Roels (1983).
more structured (two-compartment) biokinetic model is shown in Fig. 4.8

(Williams, 1967, 1975): a synthetic compartment of biomass X K , which is
mainly RNA, and a genetic-structural compartment, Xa, which consists
mainly of proteins and DNA. Models with more than three compartments for
cells are seldom needed (Harder and Roels, 1982; see also Sect. 5.7.2). Mathe-
---!'----1(G
FIGURE 4.6. Scheme of a two-compartment model for oxygen transfer in a stirred tank.
The liquid phase is structured into a mixed zone (L 1 ) and into a zone with freely rising
bubbles in analogy to a bubble column (L 2 ), with a connecting exchange (Fp) due to
impeller pumping. (From Oosterhuis, 1984.)
,,..., - --------...,
, I
~ hydrodynamics J bulkflow L I
~ arou nd the ,I
I ,
~I
stirrlZr ,
, I
I
I ,
,
I
,
I
,
,I
I
I
gas
rlZcirculation
,
,
I
,
,
I
I
, I
I
I I
gas holdup
I
I
t I
'-oxygen tranSflZr l :
,I
I I
I
I
I
I bubble I
I Hao I
diamlZter-
I
,I
,
I
I
I I
- - - - - - - - - - - - - - - - - -
. ..
L. - ------ -- ...J
~-,---
V -
FIGURE 4.7. A somewhat structured model for oxygen transfer in a stirred-tank reactor
in which only the hydrodynamics around the stirrer are taken into account. (See
Oosterhuis, 1984.)
rSK
r------------- --,
L__________________ J
FIGURE 4.8. Block diagram of a two-compartment model (K and G) of cells following
the concept of Williams (1967,1975). The rates are rSK = rate of conversion of substrate
to K compartment, rKG = rate of conversion of K to G compartment, and r GK = rate
of depolymerization of G to K compartment. (From Williams (1967) and Harder and
Roe1s (1982).)
TRANS I ENTSH
,
LAGS .... _ - - _ .........
rar(t) / '" ",
INDUCTION / REPRESSION ,
I \
\
I
DNA (gen) I
~, "transcri pt ion" I
protein synthesis
... - - - - -.... m ~NA
/" ';;ATABOLITE "'~' "translation"
I "E (ENZYME/MICROBIAL CELL)
REPRESSION
S.--------------~~~--------~~
I, SUBSTRATE INHIBITION/I j
' / ' I
_--
' ,........ . . - "
\ PRODUCT INHIBITION I
I
SUBSTRATE ACTIVATION', /
r-r(S) ..... ....- /'
ORDINARY KINETICS- -- -- --
"BALANCED & FROZEN"
FIGURE 4.9. Schematic representation of enzyme regulation principles as a basis for

biokinetic model building. While ordinary kinetics are analogous to chemical kinetics
(heterogeneous catalysis), enzyme induction and repression are typical biological
phenomena creating transients. (From Moser, 1984.)
matical models of biokinetics, thus, must include both ordinary kinetics

(substrate activation and inhibitions, in analogy to chemical kinetics) and
transients (induction and repression of enzymes), as shown in Fig. 4.9. (see
also Sect. 5.2.2.2).
For process kinetic analysis, the following rules can be given:

1. The validity of experimentally measured kinetics on any level is restricted
to the reactor type investigated. These macro kinetics should not be extra-
polated to another scale or reactor.
2. Because interactions between environment and metabolism are often
poorly understood, kinetic data and models should be transferred to other
reactors only within the range of "kinetic similarity."
3. To some extent, the simple unstructured kinetic models can be used for
describing complex bioprocessing if the parameters are adapted to process
behavior by formulating them in a time-dependent way [e.g., J.1.maX' K., and
Y = f(t)].
4. Simple unstructured models, e.g. Esener et al. (1983), can be successfully
used to describe macroscopic behavior if the model is modified to include
additional phenomena such as S inhibition, 2-S limitations, a lag phase, a
stationary phase, or a dead phase. Adaptive modeling is essential (Moser,
1981). This applies in all cases of transients, in balanced, frozen, and real
transient situations.
5. Structured modeling is needed only in cases of real transient processing with
'e ~ 'n or in cases where the purpose of the investigation is not engineering
calculations of conversion but rather a microscopic description of metabolic
mechanisms (Toda, 1981). Such models should be of more interest for
engineering when better process control can be realized through develop-
ment of more specific sensors. The reactivity of different strains is waiting
to be investigated.
4.3 Test of Pseudohomogeneity

A systematic analysis of bioprocesses must take into account the interaction
of biology and physics. With so-called "perfect reactors," pseudohomogeneity
is assumed, but this approximation should be carefully checked in advance.
Pseudo homogeneity involves quantitative tests to ensure that this condi-
tion is fulfilled. These tests, along with their quantitative criteria, are presented
in Table 4.3 for transport steps 1 through 4.
As previously indicated, the criteria of mixing time, circulation time, and
CTD should be connected in a logical way with a biological parameter to
permit a biologically relevant conclusion. The comparison of a "reaction
time," tn with a "mixing time" under conditions of a biologically determined
degree of mixing presents one poor possibility. A reaction time for a biotech-
nological process is best defined on the basis of oxygen consumption (see Table
4.1): This is one of the most rapid and easily measurable reactions (see Fig.
3.16 and Bryant, 1977; Moser, 1977c).
A quantitative test to establish that ku at the liquid-solid interface is not
a limiting factor is difficult: In general, one assumes that, with good mixing in
the liquid phase, the turbulence also has an effect on the L film at the LI S
4.3 Test of Pseudohomogeneity 147
TABLE 4.3. Quantitative tests for Pseudohomogeneity of bioprocesses as a

case of three-phase systems.
Problems I n the case of Criteria
1. GIL (kLlJ High soluble gases kG

I1T, Aero bic processes ku . a(o* - 0) 2: qo x (gas dynamics!)
d i5Ll , flocs only 0.3 :0; Ha :0; 1 --> I1T, > 1 with
Ha = ~(_2_. k . c*"-'. D)1/2
kLl n +1 '
2. L phase DCSTR and CSTR tm:o; 0.1 t,
Micromixing Recycle reactors tc :::;; tr
Macromixing CSTR N --> 1
CPFR Bo --> CD
3. LIS (kul Floes Calculate (analogy to chemical processes)
Sh L2 = 2 + 0.4 Re,1/4 . Sc l/3
Films kL2 - V L 0.7
4. S phase (D,) Flocs and films d:o; d odt

l-S limited Calculate
. =~.(l +2c*/Ks )1/2 .(Y.Ds Ks )1/2
dent K s 1 + c*/K s a max P
From Moser, 1980b.
interface. Despite this, calculations of ku can be made using either the theory
of "local isotropic turbulence" (Kolmogoroff, 1941) or the theory of "relative
velocities" between the Land S phases (Frossling, 1938). In essence, the
Kolmogorofftheory gives equal values of ku for equal power used in mixing;
the theory of relative velocities assumes that the ku value is that which reflects
the proper terminal velocity. Sano, Yamaguchi, and Adachi (1974) presented
a comparison ofthe most important literature on the LIS coefficient ku. They
showed the apparent existence of a uniform correlation between Shu and a
Reynolds number Reo based on the energy dissipation 8 (8 = mean energy
distribution or power utilization per unit mass [cm 2 /sec 3 ]). This is relevant
to stirred vessels (ST) and also bubble columns (BC):
Shu = (2 + 0.4 Re;/4. Sc l/3 ). qlc (4.1 a)
with
(4.1 b)
and
(4.1 c)
or
(4.1 d)
The factor ,pc is, after Carman, the surface factor, which is calculated from the
particle thickness, grain diameter, and specific particle surface area. With this,
an approximate value for kL2 can be estimated using Equ. 4.la. For biological
films, the dependence of kL2 on the flow velocity has been found to be 0.7 (La
Motta, 1976a). A preliminary estimate can be made of the relative importance
of the individual transport steps in GILlS process, using the concept of
effectiveness factors Yf (cf. Sect. 4.5 and Equ. 2.50).
Sylvester, Kulkarni, and Carberry (1975) introduced a strategy in which a
judicious estimate of certain physical parameters permits estimation of the
intrinsic reaction rate coefficients in GIL IS processing. An overall effectiveness
factor if is defined as
_ 1
Yf=--- (4.2)
1 + Dao
with
YfLlskrYfrcs
D a o =-----'---::---- (4.3)
kLl a
where Yfr is the intraparticle effectiveness factor (cf. Sect. 4.5.2) and YfLls is the
liquid particle effectiveness factor given by
(4.4)
with
(4.5)
Also a GIL effectiveness factor YfGIL can be defined

1
(4.6)
with
kLl
DaGIL = kG (4.7)
Substituting these equations into the full equation for the effective rate in the
S phase for a three-phase heterogeneous bioprocess at steady state
(4.8)
which results in the following equations for the total effective rate constant
k eff -_ YfLls Yfr k r Cs

(4.9)
1 + YfLls Yfr k r (CS/kLl a)
4.3 Test of Pseudohomogeneity 149
0< Tl L/S < 1
1... = _1_+_1_+_1_
-1- = -1- + -
1-
kEFF k L1 a L k L2a S ~FF kL1a L kL2a S krcllr
EO 4-11 EO 4-12
FIGURE 4.10. Flowchart of procedure for checking different transport limitations,

according to a strategy presented by Sylvester et al. (1975). For explanation see text
and Equs. 4.2 through 4.10.
and
(4.10)
Figure 4.10 summarizes the procedure of intrinsic kinetic parameter estima-
tion. The following steps are included:
1. if can be checked using Equ. 4.2, estimating kLl a and 11GL from an experi-
mental value of 'eff and a known value of He.
2. 11L[S must then be checked on the basis of above data using known correla-
tions for k L2 as (cf. Equ. 4.1a).
3. With known 11L[S, 11r kr can be calculated from DaLis using Equ. 4.5, and
then 11r from the plot 11r versus Thiele modulus rjJ (cf. Fig. 4.36).
As shown in Fig. 4.10, finally, the effective rate of individual experiments can
be interpreted as a function of different types of interactions between the
reaction and external or internal transport (cf. Equ. 4.11-4.13 in Fig. 4.10).
In principle, internal transport limitation, can be checked by measuring
the characteristic dimension of the biocatalytic mass and comparing it with
rx r--------------------------------------,
rX MAX -
15
10
--- _---In
----- -- / \
'\
5 ..... \
- ..... 1 \
r..... \
STR .......... \
10 20 30
"Xl MAX
x
FIGURE 4.11. Schematic illustration of the influence of mixing phenomena on micro bial
growth behavior creating macrokinetics: Operation conditions, for example, stirrer
speed n in stirred tank reactors (STR), or reactor construction, as in a cycle tube cyclone
reactor (CTCR) (see Ringpfeil, 1980), affect the plot of growth rate rx versus biomass
concentration x drastically, especially when x values are high.
the critical diameter, deril , which can be found in the literature (for example,
Atkinson and Daould, 1970; Kornegay and Andrews, 1968). The values vary,
but they are typically about 0.1 mm. The application of this concept ofbiofilm
kinetics to biological flocs is described in the literature (Atkinson and Ur-
Rahman, 1979). Experimental data are lacking for various types of particles
and films (Wuhrmann, 1963). Atkinson and Knight (1975) gave an estimate
for the ideal or critical film thickness for the I-S limitation; the value of Ks
and the effective diffusion coefficient in the solid phase are particularly impor-
tant factors in this calculation. The same group (Howell and Atkinson, 1976)
also proposed an extension of the method to deal with cases of 2-S limitation.
No doubt additional effort is needed in the kinetic analysis of a process to
optimize a laboratory-scale bioreactor using the criteria listed in Table 4.3.
The results, however, should justify the effort. Figure 4.11 illustrates this
problem showing that biokinetics seems to be dependent from the type of
bioreactor. Productivity is plotted for a stirred tank with increasing rotational
speed n and is compared with that in the cycle tube cyclone reactor (Ringpfeil,
1980).
Finally, it should be mentioned that experimental measurements of process
variables are made almost exclusively in the liquid phase. The assumption is
that there is at least equilibrium, ifnot identity, between concentrations in the
L and the S state. It has been shown that the correlation between external and
internal fluxes is poor (Barford and Hall, 1979).
An overview of the principal mathematical techniques for process kinetics
is given in Fig. 4.12. With real processes, the concentration terms of type 1
and type 2 are coupled with transport terms in or between fluid (GIL) and
solid phases, leading to types 4 and 5. This scheme is equally applicable
to chemical and biological reactions: Analogies to chemical kinetics often
4.4 Parameter Estimation of Kinetic Models with Bioreactors 151
i'lATHEMATI CAL i10DELS

r=f(x,k)
MICROKINETIC APPROACH MACROKINETIC APPROACH

"PROCESS KliiETICS"
FOR FOR
HOHOGENEOUS AND HETEROGENEOUS REACT! OilS HETEROGENEOUS (HULTI-PHASEl SYSTE~lS
TYPE 0 r = f1(c) TYPE 0 r = ~ (c,DJ

CONCEIHRATI ONTER~l " EXTERNAL TRANSPORT LIMITATION"
FLUID REACTIONS IN L-PHASE
FLUID/FLUID-,FLUID/SOLID-REACTIONS
TYPE r = f2 (c) TYPE 0 r = f5 (c,DS )

CONCENTRAT IOiHERM " INTERNAL TRANSPORT WlITATION"
HETEROGENEOUS REACTIONS IN S-PHASE
FLU I 0/ SOLI D- REACT IONS
TYPE0 r=f3 (T) TYPE @ r = f51c,Ds)

TE~lPERA TURE TER~1 HETEROGENEOUS m CROB I AL SYSTEr1S
"BIOFIU~KINETICS"
FIGURE 4.12. Classification of mathematical models of kinetics (types 1-5) and their
areas of application. (From Moser, 1981.)
provide useful problem-solving methods. The description in Fig. 4.12 results

from the use of the expression "homogeneous reaction" for reactions in the G
or L phases, and "heterogeneous reaction" for reactions at a solid surface
(internal or external).
4.4 Parameter Estimation of Kinetic Models with

Bioreactors
Under the assumption of pseudohomogeneity (cf. Table 4.3) and under con-
stant temperature conditions, the kinetic analysis of a process is reduced to
the search for functions 11 (c) and/or 12 (c), as specified under types 1 and 2 in
Fig. 4.12. The distinctions between reactors with or without a concentration
profile is also a decisive factor; that is, the distinction between so-called
integral and differential reactors is a necessary one.
4.4.1 INTEGRAL AND DIFFERENTIAL REACTORS
As an aid in the precise definition of integral and differential reactors, Fig.

4.13 shows a conversion versus time diagram, and the transfer of these data
to a continuous tubular reactor (Moser and Lafferty, 1976).
A differential reactor works with small concentration differences, for exam-
FIGURE 4.13. The principles of integral

C4 rr and differential reactors as shown by a
conversion/time plot ((/t), and the trans-
C3 ......... fer of the corresponding concentration
c2 data to its position along the length of
a continuous tubular reactor. (From
:
Moser and Lafferty, 1976.)
c1
:
23 4 -t=V/F
CASE. CPFR:
--Il c1
:
c2 c3
~
I-F
INTEGRAL REACTORS
/~
TRUE INTEGRAL REACTOR PSEUDO INTEGRAL REACTOR
i 1 1~F
llz
Co
~ c
...
L -_ _ _ _ _ _ _
:
;

C
ex
z
c~c.I
:
~
....
:

:
:
:
:
. cex
Z
FIGURE 4.14. Integral reactors: classification and concentration profile for a continuous
tubular reactor.
pIe C3 - C2 , and is usually defined such that the conversion, is small (rule:
, < 10%). An integral reactor is one in which the conversion is large (rule:
, > 50%) and in which large concentration differences are found, for example,
C4 - C l' A closer look at the behavior of these two reactor types is given in
Figs. 4.14 and 4.15.
For the integral reactor
t = -IC ~dc ==
Co r
Co is'' ~d'
0 r
(4.14)
follows from Equs. 2.3d and 3.94. The conversion data in an integral reactor
represent the integral value of the reaction rate operating in each volume
element of the reactor with the hypothetical concentration profile shown in
DIFFERENTIAL REACTORS
~~
TRUE DIFFERENTIAL REACTOR
PSEUDO DIfFERENTIAL REACTOR
" GRADIENTLESS REACTOR"
F----O-F
C=:
c ::
FIGURE 4.15. Differential reactors: concentration profiles for a differential reactor and
for a gradient-free (recycle) reactor.
Fig. 4.14. A special type of integral reactor (pseudo-integral reactor) is one

constructed with taps at various distances along the length so that samples
may be removed and the actual concentration profile measured. A disadvan-
tage of integral reactors is that the balance equations are a system of coupled
differential equations. The measured conversion often is due to a complex
interaction of transport and reaction processes. For quick, empirical, and
pragmatic process development, the integral reactor may be well suited,
especially now that fast digital computers and effective integration algorithms
facilitate parameterization.
Precise kinetic measurements are, however, better (made) using a differen-
tially operated reactor. A small measured change in conversion may be
directly related to a reaction rate
Co - C
r :::::; --_- 'ex - '0
= --'-"--=----=- (4.15)
t t
and transport influences or temperature gradients are of negligible conse-
quence. This approximation is permissible when the reactor volume is small
enough or when the amount of material flowing through a continuous tubular
reactor is sufficiently large. Under such a circumstance, the reactor may be
described by a system of algebraic equations, and this facilitates evaluating
the experimental measurements. For a complete series of measurements, the
reaction rate must be considered dependent on the amount of product already
present (see Fig. 4.13). A differential reactor must be primed with different
starting mixtures; this is most conveniently done by inserting an integral
reactor in front of the differential reactor. The high degree of precision
required of the analytical methods used to measure small differences in con-
version is another problem with differential reactors.
The analytical problems with differential reactors can be overcome by using
so-called gradient-free reactors; for example, loop reactors (Fig. 4.15). In this
method, after conversion in a differential reactor, a portion of the reaction

mixture is recirculated (Fr) and mixed with fresh input (F) before being reintro-
duced into the reactor (VR).
When Fr F, the concentration difference (c 1 - c) between the actual
reactor input and output is small, although the concentration difference
between the fresh input and the output (co - c) is large. This minimizes the
analytical difficulties (c in = Co and Cex = c), and
Co - c (4.16)
The conversion per reaction cycle is a differential amount, but It is at the high
level of conversion typical of an integral reactor. The loop reactor is therefore
referred to as a "pseudo-differential reactor" or "gradient-free reactor." The
identity indicated in Equ. 4.16 can be more easily recognized when the two
parts that result from the mass equivalence of a recycling reactor (left side)
and set equal to that of an ideal stirred vessel (right side), and the resulting
equation is then solved for C 1 :
(4.17)
Only for Fr F does C 1 ~ c (Hoffmann, 1975). At Fr;;::: 100 F, the recycling

reactor is kinetically equivalent to the CSTR (Luft and Herbertz, 1969), which
is the type most often used for the kinetic analysis of homogeneous, liquid
phase reactions. The recycling reactor is most often used for kinetic measure-
ments involving a solid phase (heterogeneous catalysis); due to the recycled
stream there is high fluid velocity, which results in good transport properties
at the LIS interface (see Equ. 4.1a). The recycling reactor combines the advan-
tages of differential reactors (specifically, that the reaction rate can be obtained
directly from experimental data and related directly to conditions and concen-
trations) with the advantages of integral reactors (large differences in input
and output concentrations that are easily measured). The different reactor
behaviors require individually adapted methods for evaluating the kinetic
data, and these are now to be discussed.
4.4.2. INTEGRAL AND DIFFERENTIAL REACTOR DATA

EVALUATION METHODS
The connection between integral and differential reactors and that between
integral and differential methods of data evaluations are shown in Fig. 4.16
(after Froment, 1975). Data from integral reactors can be evaluated in the
same way as data from differential reactors if the data are first numerically
differentiated, or differentiated analytically or, more often, graphically. In
cases where integral data are to be evaluated differentially, the following steps
should be followed:
FIGURE 4.16. Relationship be-

tween differential and integral ,------------,
REACTORS
,----------,
reactors, and between differen- DIFFERENTIAL INTEGRAL
tial and integral methods for grad ientl ess with gradients
evaluating kinetic parameters. 1. cSTR 1. cPFR
t"high
2. recycle R 2. botch R
~ F
3. cPFR
" ~Iow
r= c=co r;!:c~:co
t t
I differenti~ ada~t by
'''''1/
DIFFERENTIAL
'TO'"'
INTEGRAL
KINETIC EVALUATION
TABLE 4.4. General definition of rates in different reactor operation modes.

Reactor operation
(T = constant) Profile Rate, ri Equ. No.
dc,
DCSTR (V = constant) f(t) r=- 2.3d
, dt
(2.4a-c)
dN
DCSTR (V = variable), SCSTR f(t)
r, = V(t).dt Vet) = Vo + F(t)t
3.86
(VR + 1I;)dc,
DCSTR + recycle f(t) rj = 3.100;
VRdt
if 11; VR then
2.3d
_ dc,
CPFR fez) r, = vZdz 3.94
F
CSTR (V = constant) ri = -y(cex - c in ) 3.87
(F+F,)
CSTR + recycle r, = ---V-(c" - c'n) 3.1 01;
if F, F then
3.87
1. Draw the concentration versus time curve.

2. Draw a smooth curve through the measured data points.
3. Determine the slope of the curve using numerical, analytical, or graphical
methods: e.g., (dcJdt) = rj for DCSTR. Table 4.4 lists the definitions of
"reaction rates" in different reactor operations.
de
lIT EXPER IMENT
f de
t _~10DEL f(e)
EXPERlllEr<T
0
" k
i t k
" 'IIOOEL
0
i
0
--f(e) _t
4.17 4.18
FIGURE 4.17. General plot showing the differential method of evaluating experimental
data to obtain the parameter k of the kinetic equation r = k J(c).
FIGURE 4.18. General plot showing the integral method of evaluating experimental
data to obtain the kinetic parameter k.
With measurements made in a differential reactor, the values of rj are

obtained directly. The important step of a differential data evaluation pro-
cedure consists of a comparison between the experimental data and the hy-
pothesis of Equ. 2.53. Usually, this is done graphically using a linearization
procedure (see Sect. 2.4), as is shown in Fig. 4.17 for the general case. The
function f(c) represents the mathematical function that was postulated in
constructing the model. With the computer, nonlinear regression methods are
also a realistic alternative.
In an integral data evaluation scheme using data from integral reactors, the
following steps are included:
1. Fit the function representing the hypothesis (Equ. 2.53) to the experiment
using analytical or graphical techniques so that
f~ C
Co
"
f(c)
= kf' dt
0
(4.18)
2. Compare the experimental data with the prediction that follows from the
hypothesis; this may often be done using the linearization shown in Fig.
4.18.
If the agreement between the hypothesis and the experimental data is
satisfactory (by statistical tests), the kinetic parameters can be taken from the
slope of the straight line and, in some cases, from the intercepts with the axes.
The correct model must be identified before parameterization. This procedure
for the analysis of process kinetic data is a general one and has been success-
fully used for chemical processes also (Levenspiel, 1972).
The application of the procedure to bioprocesses should be considered an
analogy. Limitations on the validity are to be expected due to the adaptability
TABLE 4.5. Bioreactor operation techniques and biological behavior.

Reactor Biological
operation growth "Balanced" "V nbalanced"
Gradient -free Steady state CSTR
Quasi-steady state "Extended culture"
Unsteady state ~--- CSTR,SCSTR ---~
"Periodic" ~-- "Transient operation techniques" --~
With gradients Steady state ~--- CPFR, NCSTR ---->
Unsteady state ~-- DCSTR, CPFR, NCSTR --->

"Periodic" ~-- "Transient operation techniques" --->
From Moser, 1980b.
of biological materials: Organisms are able to adapt themselves to changes in

their environment by altering their metabolism (cf. Sect. 4.2). Therefore, a
criterion should exist for partitioning a process into reactor operations and
biological factors. The concept of "balanced growth" is used in this way, and
the organism lives in completely compensated conditions. Table 4.5 opposes
bioreactor operation and biological behavior (Moser, 1980b). A narrow classi-
fication system is not yet possible due to lack of relevant data. A factor that
may be significant is the relative behavior of the time constants (kinetic rate
constants) between a disturbance of the external environment and a distur-
bance of the inner environment of the biological organism (as outlined in Sect.
4.2 as a case of regime analysis).
There is a difference in the use for process development of the different
kinetic parameters obtained from the two calculation methods, each referring
to a different process operating technique. The kinetic parameters for the
various process operation techniques are primarily bound to the special type
of reactor system investigated through the coupling indicated in Table 4.4
(Moser, 1978). The values can be extrapolated to other reactor systems only
to a first approximation. However, Esener et al. (1983) have recently pointed
out that under certain circumstances no significant difference exists between
the parameter values obtained in different modes of cultivation.
4.4.3 RESULTS OF DIFFERENTIAL AND INTEGRAL ANALYSIS:

LINEARIZATION DIAGRAMS
Since kinetic data from a series of DCSTR experiments not only serve as the
basis for process development in the DCSTR but also serve to a first approxi-
mation for the design of continuous reactors, integral and differential data
evaluation methods for the case of a DCSTR will be considered in more detail.
Figure 4.19 shows the contrasting concentration/time curves for a discon-
tinuous fermentation (Moser and Lafferty, 1976). For integral analysis, the
typical procedures are represented as if carried out in a microbiological
laboratory using shaking flasks. A series of i experiments is carried out, each
(PSEUDO) DIFFERENTIAL METHOD INTEGRAL METHOD

~ - 0 with LI t - 0
~
i-1
SOi"
, ,
-2
3
4
"-
t
,u=f(SO .)
,I
FIGURE 4.19. Differential and integral methods for the kinetic analysis of a process in
a discontinuous reactor with microbial floes. (From Moser and Lafferty, 1976.)
with a different starting concentration of substrate, So. In each trial only the
cell concentration, x, requires measurement. The data evaluation follows from
Equ. 4.18. The integration of the function representing the hypothesis, Equ.
2.5a
fXo
x dx
-=f.1
x
it
to
dt (4.l9a)
results in
(4.l9b)
By means of a semilogarithmic plot of the values over the range t2 to t1> f.1
may be obtained
(4. 19c)
As a result of the integral evaluation one finds that, for example in the 1)1odel
identification of Monod kinetics (Equ. 2.54), f.1 can be given as a function of
the initial concentrations SO,i
(4.20)
In contrast, with the differential method of analysis, only a single experiment
need be carried out. However, in this experiment x(t) and set) are simultane-
ously measured. As suggested in Fig. 4.19, measurements made at about
2-hour intervals are sufficient from the beginning of the experiment to the
beginning of the exponential growth phase. In the range where Monod kine-
tics are to be characterized in effect, measurements should be repeated every
few minutes (this is also the reason why, for example, semicontinuous addition
methods make possible more satisfactory kinetic measurements; Reuss, 1976;

Webster, 1981). As previously indicated, calculation of the rate of growth, J.l,
is done directly from the slope of the x/t curve by graphical differentiation
over a particular range of L1t:
1 L1x
J.l=_.- (4.21)
x L1t
where x is the average value of x in the measured interval L1t.
The result of a differential analysis of a DCSTR process is a relationship
giving J.l as a function of s, which is itself a variable changing over the course
of the reaction process
J.l = f(s(t)) (4.22)
In this connection it might be mentioned that Monod (1942) originally regarded
the relationship that he proposed for microbial growth kinetics (based on an
analogy to enzyme kinetics) as equally valid for continuous and discontinuous
processes. One was here dealing with formal kinetics and a case of parameter
fitting (Ks instead of Km).
Finally, briefly presented here are the results of a differential or integral
analysis of the various kinetic formulas shown in Fig. 4.12. The solutions are
also useful as analogies to many processes in chemical, biological, and food
technology.
Figure 4.20 gives a comparison of the results for type 1 kinetics: the so-called
"power law equation," which operates with the concept of the reaction order,
n. The diagram allows determination of n and the rate constant k. A number
of variants are to be found in the literature (Levenspiel, 1972). In this case, the
TYPE 1 , r = k.c n
DIFFERENTIAL ANALYSIS I NTEGRAL ANALYSIS
Ig r = Ig k + n . Ig c =(n-1)k.t
--n:1-~
c Co
Ig r 1
cn- 1
~ln-1Jk
jl9k j n1_1
L-~cO~___________ t
~-------------Igc
n" 1
FIGURE 4.20. Determination of the parameters of a type 1 kinetic model (Fig. 4.12) using
differential or integral methods.
integral evaluation requires a "trial and error" method where nand k should
be obtained simultaneously:
1 1
n-l - n=T = (n - l)kt (4.23)
C Co
This method may be simplified to a graphical presentation when one simply

assumes in turn that n = 0, 1,2, and so on.
In contrast, the differential evaluation of the hypothesis r = k c n gives
logr = logk + n'logc (4.24)
with the appropriate graphical presentation, Fig. 4.20.
The type 2 kinetics of Fig. 4.12, with the general form
k1c
r=---- (4.25)
1 1 + kz ' c
is applicable to several different cases:

1. Homogeneous chemical reactions in which the order of the reaction changes
2. Heterogeneous chemical catalysis reactions following the Langmuir adsorp-
tion isotherm relationship with kl = k K, k z = K, and K = adsorption
equilibrium constant, k = rate constant for adsorption which, when multi-
plied by the occupancy factor (i.e., fraction of surface covered by adsorbed
(molecules), gives the rate of adsorption (cf. Equ. 2.56)
3. Enzymatic reactions, where rmax = kdkz and Km = llk z (cf. Equ. 2.54)
4. Microbiological reactions, where rj ~ J1 (or qJ, c = s, and kllk2 ~ flmax (or
qi.max), Ks = llk z
Figure 4.21 shows the result of the integral analysis of type 2 kinetics in
graphical form. For part (a), after the integration of Equ. 4.25, one obtains
lncolc = -k2 + ~ (4.26)

Co - c Co - c
For part (b) of the figure, after integrating Equ. 2.54 one obtains the so-called
"Henri equation," that is, utilizing the nomenclature for enzyme reactions
(4.27)
Multiplying the Henri equation by lit makes possible a linear, graphical

represen ta ti on:
So - S 1 So
. _ - = -K.-ln-
t S t s + rmax (4.28)
This method is known as the "Walker plot" method (Walker and Schmidt,
1944), and it is often used in the area of biological waste water treatment
(Wilderer, 1976). The Walker plot has the advantage that, for various types
of kinetic models (power law equations with different reaction orders as well
k1 . c
TYPE 2: r =---'---
1 k 2 .c
50 -5 1 5
--=-K -Ig~
t 5 t 5
50- 5
-t-
n -1
1 n=2
I I
I I
I I
rmax - - - - - ~---- n =0
..../1
/ 1
a ,,/ 1
I
.11g 5 50 b
------t
FIGURE 4.21. Determination of the parameters of a type 2 kinetic model (Fig. 4.12)
using integral data evaluation methods. In (b), the "Walker diagram" shows a corre-
sponding integral evaluation method for various kinetic equations. (Adapted from
Wilderer, 1976.)
as type 2 kinetics), straight lines of differing slopes are obtained, as indicated

in Fig. 4.21b.
Strictly considered, the Walker plot applies to reactions in which only
substrate is consumed; the growth of cells is not anticipated and there is no
other reaction. This linearization starts from the enzyme kinetic equation
ds s
- dt
- -,.,s
-
=,.,S,max Km + S (2.54)
which differs from the equation used in microbial kinetics for the rate of use
of substrate
ds s
- dt = rs = qs,max K. + s . x (4.29)
The connection between enzyme and microbial growth kinetics is

rS,max = qS,max . X (4.30)
which is also reflected in the dimensions ofrs,max [mg/l' min] and qs,max [h- 1 ].
The Walker plot is, therefore, a special case with dx/dt = 0, that is, with
YxlS = O.
A complete description of microbial reactions includes at least two differen-
tial equations: one for substrate (S and/or 02) and the other for growth. The
system of equations for the reaction scheme of Equ. 2.1 is, in the simplest case
dx S
dt = ILmax Ks + S. x (4.31)
ds S
- dt = qS,max Ks + S x (4.32)
do 0
- dt -- q O,max Ko + 0
'X
.
(4.33)
The solution of this system of coupled differential equations for the variables
X and S (or 02) is complicated, and it is done primarily using the concept of
the yield coefficient, Y, to eliminate the variable S or 02' Equation 2.13, written
in another form, is
(2.13)
The complete solution with determination of the kinetic parameters in this
case is, however, still a graphical trial and error procedure, linearizations being
used only in secondary plots using intercepts and slopes of the primary plot
of, for example, lit 'In (xlxo) versus lit 'In(x max - xolxmax - x) for growth
data (Gates and Marlar, 1968). Instead of the Henri equation, the following
equation is found after integrating Equ. 4.32 (using Equ. 2.13 for substrate
consumption)
qS,max' x max ' t =

So
Ksln- + (1
- ' Xmax + Ks) In-----'--
xmax - Y
x1s ' S (4.34)
S Yx1s Xo
Equation 4.27 is a special case of this equation: It can be obtained as the
limiting case when Yxis --> 0 in the second term of the right side of Equ. 4.34.
Linearization of Equ. 4.34 is possible by multiplying by lit, resulting in
~ In Xmax - Yxis ' S = qS,max' Yx1s ' Xmax _ Yxis ' Ks . ~ In So (4.35)
Xo Xmax + Yx1s ' Ks Xmax + y. Ks t S
In contrast to the Walker plot method, in the Gates linearization method
the curves for various conditions (so, XmaX' and xo) are not superimposed; they
vary depending on the conditions. An alternative Gates diagram can be drawn
using a graphical trial and error method and Equ. 4.36
~ln~ = c[ln(l + ad)] _ b (4.36a-e)

t So t
b= !lmax (xo + ~Is' so)

~ls'Ks
Xo + ~Is'so
c = 1 + ---'----
~lsKs
d = So - s
FIGURE 4.22. Batch process evalua-

tion using the integrated form ofbio-
kinetics (see Equs. 4.36a-e) with the \
\
aid of a graphical trial and error
}~.,
method.
b
1
L...-_-'--_-'--_-'-_-'-- tin (1 + ad)
Plotting lit In(slso) versus [In(l + ad)jt] by assuming values for a, and iterat-
ing this procedure untillint:arization appears, leads to the estimation of kinetic
parameters (see Figs. 4.22 and 4.23). A variant of this procedure, which also
starts from the integrated form of the Monod equation, is used for determining
Ks for a discontinuous process from the position of the inflection point of the
growth curve (Meyrath and Bayer, 1973).
This short presentation of integral solution techniques for biokinetic equa-
tions shows that integration of the function incorporating the hypothesis
works only in the simplest cases. In addition, the integrated form lacks much
of the flexibility desired for fitting data; there is an immediate tendency to
resort to descriptive polynomial functions. One also tries to use simplified
forms of enzyme kinetics such as Monod kinetics, and this involves "power
law"-type equations (cf. type 1 in Fig. 4.12 or Equ. 2.2a).
For the case of low substrate concentration, such as in biological waste
water treatment, one can ultilize a formal first-order equation. From Equ. 4.29
ds J1.max s
--=----x (4.37)
dt Y Ks
and with x = Xo = constant, the integration can easily be carried out to give
In So = ( J1.max ) xot (4.38)

s Y'K s
so that using a semilogarithmic plot, the factor in parentheses may be obtained
from the slope.
The missing value for Y must be calculated for the region So Ks using an
equation that is formally zero order with respect to S. With J1. = J1.max, inte-
grating Equ. 2.5a one obtains
(4.39)
and substituting this in a kinetic equation of zero order in S
ds J1.max
--=--x (4.40)
dt Y
1
tl" xo/x
/ INTERCEPT a
1 X MAX - Xo
~----~-------------------------------I"-=~~
t X MAX - X
SLOPE
~--------------------------------------XMAX/YX/S
SLOPE
c
L.-_________________________________ INTERCEPT
FIGURE 4.23. Alternative integral method following linearization according to Gates

and Marlar (1968) on the basis of a primary plot (a) and secondary plots using the
slopes and the intercepts of the primary plot (b and c).
one obtains on substitution and integration

s - s 1
_0_ _ = _(el'm,,t - 1) (4.41 )
Xo y
The corresponding plot gives a straight line with slope l/Y.
In general, it is apparent that one can use integral data evaluation methods
only with the simplest types of kinetic equations. When additional kinetic
effects appear (see Chap. 5), as they always do in real processes, the integration
step is more difficult or is impossible.
For this reason, in complex cases the differential data evaluation method
is used. The differential analysis of DCSTR is of the greatest significance:
The DCSTR is the type most often used in practical situations. Also, for
developing processes for unconventional bioreactors (for example, bubble
columns, towers, and tubular reactors), it is significant that these all have
integral reactor behavior.
There are several alternatives to a graphical solution for the type 2 kinetic
equations (Fig. 4.12) that form the basic type of kinetic model. In Fig. 4.24c,
the commonly used Lineweaver-Burk (1934) linearization is presented by
Equ.4.42:
1 Ks 1 1
-=~.-+-- (4.42)
11 I1max s I1max
The full procedure is shown from experimental data manipulation in Fig.

4.24a, with experimental errors to the setup of a Monod diagram in Fig. 4.24b.
Significant distortions result from the double reciprocal plot and the concen-
tration of data in a certain region (points 2-7).
Due to the distortion in the treatment of errors associated with the double
reciprocal plot, it is useful to look for alternatives. A simple reciprocal plot
due to Eadie (1942) and Hofstee (1952) and corresponding to Equ. 4.43
11 I1m.x 11
(4.43)
s
is shown in Fig. 4.25, along with the same error range of 5% (that is, the
95% confidence interval). The most satisfactory treatment is, however, the one
proposed by Langmuir (1918) for heterogeneously catalyzed chemical reac-
tions. The general form of this linearization is
c 1 k2
=-+_c (4.44)
r kl kl
which may be written for Monod kinetics as
s Ks S
-=~+-- (4.45)
11 I1m.x I1max
This linearization is shown in Fig. 4.26, again with a 5% error range. Here
the error distortion is minimal.
A completely new method of determining enzyme kinetic parameters is one
that may be called "direct linearization." It is due to Eisenthal and Cornish-
Bowden (1974) and is based on the direct dependence of 11 on S, leading to the
dependence of I1max on Ks in the form
I1max = 11 + 11-Ks
S
(4.46)
c
-J.l= -
J.lmax 1fJ
t fJ
s
---
KS KS
J.lmax ,-K S
r "-
"- ....
"-
"- -.. -.. ....
.... "-
",
a "- -.. .....
J.l
--t \
\
I
..............
...... -
t
fJmax/Ks
fJ/ s
4.25
KS_ +_S_
2.... =__
fJ fJ max fJmax
b S/fJ
--s
V,u I
/
\
/ /......
' 1/fJ max
t
/ /
/ / t
/
I t /
~
~
/
/ KS/,umax / ~
~ /
/
..,/~
/
/
/ /
/
~
/
/
/
I/~max
t
S
c
-IlKS --1/s -Ks
4.24 4.26
FIGURE 4.24. Growth and substrate utilization curves versus time for a discontinuous
process with measured data 1-10. (a) Measured data in a cit diagram. (b) Data shown
in form of Monod plot, (---) indicates 95% confidence interval. (c) Data shown in
form of Lineweaver-Burk (1934) double reciprocal plot.
FIGURE 4.25. Model identification and parameter estimation of the Monod kinetic
equation using differential methods and a simple reciprocal plot (Eadie, 1942; Hofstee,
1952). (---), 95% confidence interval.
FIGURE 4.26. Model identification and parameter estimation of the Monod kinetic
equation using differential methods following the Langmuir (1918) approach. (---),
95% confidence interval. The deviation is minimum.
These authors showed that if the experimental S values are plotted on a

negative horizontal axis and the observed r values are plotted on a vertical
axis, then straight lines drawn through the corresponding -.8 and r points
intersect at 8 = Km and r = rmax.
While Fig. 4.27b represents the conventional r = /(8) plot (see also Fig.
4.24b), Fig. 4.27a demonstrates the simple graphical procedure for estimat-
rmax RESP r
a A b
6 5 4
- - KmRESP 5
FIGURE 4.27. Parameter estimation without model identification of saturation-type

kinetics (enzyme and Monod kinetics) using the "direct linear plot" (Eisenthal and
Cornish-Bowden, 1974) ofr versus s or rmax versus Km (a) according to Equ. 4.46. The
right-side graph (b) shows the conventional plot of saturation-type kinetics, for better
comprehensiveness. (Reprinted by permission from Biochem. Journal vol. 139, p. 715,
copyright (c) 1974. The Biochemical Society, London.)
ing enzyme kinetic constants, which is in agreement with the rearranged

Michaelis- Menten equation as given in Equ. 4.46.
For any values of sand r it is possible to plot rmax against Km as a straight
line with slope rls, intercept -s on the Km axis, and intercept r on the rmax
axis. The most obvious advantage of this "direct linear plot" is that it requires
no calculation.
Finally, the values of kinetic model parameters can also be estimated using
nonlinear regression methods or following a direct method of applying a
computer for the simulation and comparison with experimental data (e.g.,
Aiba, 1978). Because analytical solutions of differential equations are rarely
possible, numerical solutions are preferred. The Runge-Kutta method is used:
Normally it is available from computer program libraries and is used together
with the maximum principle of Pontryagin for optimization (e.g., Bojarinow
and Kafarow, 1972; Fan and Wang, 1968).
Logically, the accuracy of parameter estimation using this simulation
method depends on the quality of the model function chosen. An inadequate
model results in pseudo kinetic parameters. The fitting of simulation to experi-
mental data is carried out with the aid of statistical criteria such as the F -test,
minimizing deviations ul:
(4.47a)
Multiresponse analysis or joint analysis uses the concept
(4.47b)
With real biotechnical processes, almost without exception one uses differen-
tial data evaluation methods (see Chap. 5).
4.5 Modeling Heterogeneous Processes

In this section the modeling procedures that can be used for a true hetero-
geneous system are briefly outlined.
Figure 4.28 presents a concentration profile of, as an example, oxygen in a
three-phase process. The transport steps that could be rate limiting are shown.
These steps involve the mass transport coefficients kG and kLl at the GIL
interface and kL2 at the LIS interface, or ks as a measure of a critical film
thickness <5s,crit above which the oxgyen concentration sinks below a critical
value, 0crit. These types of concentration profiles are typical of multiphase
processes, and they provide the starting point for the derivation of the model
equations. Apparently, the relative proportions of the various phases playa
major role in deciding whether one can work with a homogeneous model or
must work with a heterogeneous model system. Changes in the relative
velocities between the liquid and solid phase (v rel ) result in changes in the
concentration profile in the solid phase (curves a-c, Fig. 4.28).
It is possible to apply the concept of pseudohomogeneity in cases where the
diameter of small particles d p is slightly less than the film thickness <5, which,
in the two-film theory, corresponds to the transport resistance, This situation
is illustrated in Fig. 4,29. In this example of a three-phase process, the S phase
particles are suspended in the L phase. Five different concentration profiles
are shown at the GIL interface (cf. curves a-e) for various ratios of the reaction
rate coefficient kr to the transport coefficient kTR . In the case of a LIS reaction
with a large dp , only profiles a-c apply (cf. Fig. 4.28).
G
I >I<
'IOL
L
GAS LIQUID SOLID
1 I eRIT ....1
I z
I
6G 6L1 6 L2 6S,c 6S,b
MODEL
MEASURE
,
..... kG kL1
OTR
l
t m , tc
kL2
Ref
I
,
kS
ds
FIGURE 4.28. Heterogeneous three-phase model for a gas-liquid-solid process with the
oxygen concentration profile across the different films of thickness 15 (t5G , t5 Ll , t5Ll , t5s),
which are, according to two-film theory, linked to the mass transport coefficient
k (kG, k Ll , k Ll , k s ).
4.5 Modeling Heterogeneous Processes 169
L-FILM L-PHASE
III
III _ S-PHASE
41
III 0L - 0
L---~~t--~~~~~=======-----z
6
L1
'-
FIGURE 4.29. Pseudo homogeneous model approach for a heterogeneous GILlS

process. The solid phase, S, is homogeneously suspended in the liquid phase as small
particles (flocs with ds < 6Ll). It may affect mass transport through the GIL interface,
as shown in curve a-e in case of, for example, a reaction occurring in the liquid phase
(or inside the flocs) consuming 02.
All of these cases of coupling between kinetics (k r) and transport (kTR)

phenomena can be handled and solved with a uniform scheme, which is shown
in Fig. 4.30 with Equs. 4.48 through 4.59. One uses the concept of an effective-
ness factor, 1], which is regarded as a function of the ratio kr/kTR . In individual
regions of applying the 1] concept, individual, traditional descriptions of
the relationships between kr and kTR have established themselves and will be
discussed hereinafter (Thiele moduli, rPext and rPint; Damkoehler number, Dan,
Hatta number, Ha).
For model building when the problem is one of "mass transport with
simultaneous reaction," the conservation of mass equation (Equ. 2.3a-c) is
again utilized. In the one-dimensional stationary case it has the form
d2 e
D--r=O (4.60)
dz 2
The boundary conditions for fluid (gas or liquid)-solid processes are
e = e* for z = 0 (4.61a)
and
de
--= 0 for z = 6 (4.61b)
dz
In the case of fluid-fluid processes
e = e* for z = 0 (4.61c)
and equilibrium exists between on transport and off reactions for z = 6.
General ref! = f(k,. k TR ) resp. {'(t r tTR or t e ) Equ.4.48

problem:
Special
~~
Extracellular Intracellu lar
situation: level level
{'(t r t e )
~
See Sect. 5.7
Transport Transport and Sect. 2.4
limitation enhancement
General
~ Equ.4.50
r eff = rideal 1/ r Equ. 4.49
concept:
ref! neff
Definition 1/ r = - - Equ. 4.51 1/ TR =-;;--:= EOTR Equ.4.52
of 1/: r ideal ideal
kr kTR
Interpretation 1/ = f- Equ 453 1/ =f- Equ.4.54
/~.
TR kr
of"
Special External Internal

case:
_
1/r,ext -
kr
f k Equ.
1/ r, int = f( J Dan Equ.4.55 1/TR = f(Ha) Equ.4.56
TR 4.57 resp. int)
1/ r, ext = f(ext) O~1/r~l Equ.4.58 1/TR;;" 1 Equ.4.59
~ f ~
See Sect. 4.5.1 See Sect. 4.5.2 See Sect. 4.5.4
\ )
V"
Combined effects: See Sect. 4.5.3
FIGURE 4.30. Quantification and modeling of the problem of reaction and mass transfer
interaction in microbial systems: general concept for the solution.
The following general catagories characterize the interaction between kine-

tic and transport considerations:
1. Transport limitations between fluid-fluid and fluid-solid processes

a. "External" transport limitations at the GIL and LIS interface (cf. Sect.
4.5.1)
b. "Internal" transport limitations within the S phase (heterogeneous
chemical catalysis and heterogeneous biological catalysis, cf. Sect. 4.5.2)
2. Transport enhancement only at the GIL and LIL interfaces of (pseudo)
homogeneous systems (cf. Sect. 4.5.4)
The concentration profiles a through c in Figs. 4.28 and 4.29 reflect trans-
port-limiting processes; the profIles d and e in Fig. 4.29 reflect the enhance-
ment of transport due to a reaction.
4.5.1 EXTERNAL TRANSPORT LIMITATIONS

With curve a of Fig. 4.29, the reaction is unimpeded (c --+ c*); here the reaction
is under kinetic control and is in the "kinetic regime." With curve c, transport
is so slow in relation to the rate of reaction that the concentration in the L
phase (or S phase) decreases to zero; the reaction here is in the regime of
"diffusion control" and transport limitation is paramount.
Curve b is an intermediate case in which both phenomena are at work. The
equation valid for a so-called "effective" reaction rate, r eff , is (for the case of
a first-order reaction, Fitzer and Fritz, 1975) at steady state
c*
(4.62a)
reff = (l/k L A) + (l/k r V)
The reciprocal rate constants are additive in this case in which transport and
reaction steps operate in series.
One may take the ratio of the kinetic (k r) to transport (kTR' Equ. 2.2b) rate
constants as a criterion for determining the boundary between the different
types reflected in curves a and c. If(krlkTr) 1, then the reaction is "very slow"
in relationship to transport, and Equ. 4.61 may be reduced for the kinetic
regime to
(4.62b)
which is the ideal rate of reaction. If, on the other hand, (krlkTR) 1, then,
although the rate of the reaction may be slow, the transport process is even
slower. One thus has transport limitation, and Equ. 4.62a is reduced to the
already well-known equation for physical adsorption. Due to the reaction in
the liquid phase, CL --+ O.
To obtained undistorted kinetic parameters in cases such as these, transport
processes must be arranged well enough so that Equ. 4.50 applies. This is
accomplished mainly by good mixing in the fluid phase: With good turbulence,
a high kL value is attained. If Equ. 4.62a is recast in the form, for example,
1
r eff = k r V c* 1 + krlkTR (4.62c)
clearly this equation is of the same type as that found in all other cases of the
interaction of kinetic and transport factors (cf. Fig. 4.30). The solution is Equ.
4.57 resp. Equ. 4.49. A comparison of Equs. 4.57 and 4.62c shows that Equ.
4.62c reflects rideal (the ideal case, without transport limitation). The effective-
ness factor '1" defined in Equ. 4.51, is often found in the range
o : : ; '1r ::::;; 1 (4.58)
in case of external and/or internal transport limitation, and it is only greater

than unity in the case of complex kinetics or nonisothermal conditions.
The definition of'1r in the case of a purely external limitation is based on a
substrate modulus (Thiele modulus), lext, given by Horvath and Engasser
(1974) as
rmax
lext = K .k (4.63)
S L2
in dimensionless form.
For simple saturation-type kinetics, Equ. 2.54, at steady state one has
r max '
K -
S
= k L2 (SL - st) (4.64)
s+s
where SL and st are the liquid bulk and surface concentration of substrate,
which can be calculated using Henry distribution coefficient st = /He. Ss
A graphical representation ofEqu. 4.64 can easily be given using dimension-
less quantities like lext (see Equ. 4.63) and is shown in Fig. 4.31 for different
values of lext.
Fig. 4.32 represents the dependence of the effectiveness factor '1r on the
modulus lext at different values of s/K s. One observes the kinetic regime with
'1r -+ 1 and the diffusion regime, which approaches a limiting value of llimit
1
rAimit = 1 + rPext (4.65)
for reactions that are first order in substrate. Under these conditions, the
apparent first-order rate constant k 1 app is given by
rmax
kl,app = Ks . llimit = Ks(1 + lext) (4.66)
Similarly results have been reported by Toda (1975) examining the depen-
dence of pseudokinetic parameters rmax.app and Ks.app on fluid flow rate. The
following equation was found to be valid for describing the dependence of Ks
on kL2 a
Ks e
~=I+-..A.
KS 'f'ext (4.67)
as
where ep is the volume fraction of enzymic agar particles in a packed column
of immobilized cells.
The conventional method of estimation of kinetic parameters is a plot of
so(s versus In(1 - (s) according to
so' ( = Ks. app ln(1 - (s) + rmax.app t (4.68)
This is valid for plug flow reactors such as packed beds (Kobayashi and
Moo-Young, 1973; Lilly, et aI., 1966). Varying the liquid flow rate should give
T'Jr,EXT 1 r-1lll:i:E~r-~:::"",";r--",,"~=--,
A
Ql
0.01
1000
20 40 _ s/Ks . . Cl>EXT
4.31 4.32
FIGURE 4.31. Plot of normalized effective reaction rate rerrlrmax against dimensionless
substrate bulk concentration s/Ks for different values of modulus for external transport
limitation ifJext in the case of Monod-type kinetics. (From Horvath and Engasser, 1974.)
FIGURE 4.32. Plots of effectiveness factor of reaction '1"ext versus modulus ifJext in case
of external transport limitation as a function of substrate concentration s/Ks . The
rate-determining steps (rds) are indicated with their range of validity by dotted lines,
together with the limiting first-order effectiveness factor tAimit at low s values. (From
Horvath and Engasser, 1974.)
the value of Ks unaffected by external transport. A more accurate method was

presented by Lee and Ryu (1979), by dividing Equ. 4.68 by t. This yields the
Walker plot (cr. Fig. 4.21 b), which allows a more accurate estimation by means
of isoconcentration lines.
A new graphical method of the estimation of Km values (Ks) free from
external transport was proposed by Seong et al. (1981). The intercept at the
ordinate obtained by the straight-line extrapolation of data points in the plot
of Km,app values versus the reciprocal of superficial velocity in column permits
an easy and accurate calculation of Km.
The influence of external transport on kinetic parameter estimation can also
be illustrated in an Eadie-Hofstee plot, as shown in Fig. 4.33 (Hartmeier, 1972;
Horvath and Engasser, 1974). Significant departures from linearity, however,
are observed with increasing external transport limitation (,pexI > 0.1), partic-
ularly when a wide range of s is examined.
Atkinson and Ur-Rahman (1979) used an approach based on Equ. 4.62a,
replacing the rate constant kr by an equation for kr,app, which also takes into
account internal transport limitation (I'/r,iol; see next section).
With
' _ kL2
kL2 (4.69)
-
psas
they concluded that if
FIGURE 4.33. Eadie-Hofstee-type plots in

dimensionless form for different values of 101
the modulus <Pexl in case of external trans-

port limitation with Monod-type kinetics.
(From Horvath and Engasser, 1974.)
FIGURE 4.34. Sherwood number ShL2 required for a minimal liquid-phase transport
limitation at LIS interface (c\2) as a function of substrate concentration s/Ks, with
varied values of solid-phase transport limitation represented by k3' which is propor-
tional to particle diameter. The limiting ShL2 number (Sh L2 ' (l(2/y = 2) is indicated.
(From Atkinson and Ur-Rahman, 1979.)
k~2 > 10 (4.70)

kr,app -
the liquid-phase diffusion limitation will have only a small effect on the overall
rate of substrate uptake. Finally, these authors derived an equation using Equ.
4.69 and 4.70 together with kr,app, leading to the following expression
Sh 1X2 _ 60 . l1r[k 3 'IX' (l/as)]2
(4.71)
L2-Y- l+s/Ks
which is plotted in Fig. 4.34 in terms of ShL2 ' (1X2/y) versus s/K s for various
values of k3 . IX as, k3 being a coefficient of the biological rate equation repre-
senting solid-phase diffusion limitation (cf. Equ. 5.248). The curves represent
the value of ShL2 required to achieve a minimum L-phase diffusionallimita-
tion for bioflocs of given size (k3 . as) exposed to a given concentration. Clearly
the larger the particle and the smaller the concentration, the larger the kL2
value required. Since y, relating diffusivity within biomass Deff to that of the
free solution D, is probably only slightly less than unity, and IX, which is a
shape parameter, is only slightly greater than unity (1.06 for cylindrical and
1.16 for spherical flocs), it follows that the smallest value for ShL2 (1X2/y) is
approximately 2, sufficient only for the smallest floes, as shown in Fig. 4.34.
Substituting this value into Equ. 4.71 for the conditions most sensitive to a kL2
limitation, that is c -* 0, on the assumption that (with ds = lias) k3 . rJ.' ds < 0.5
(1]r'" 1), leads to the criterion that there will be no kL2 limitation even under
quiescent conditions (Atkinson and Ur-Rahman, 1979):
(4.72)
In the case of a vertical rotating drum biofilm reactor, La Motta (1976a)
measured the kL2 value and correlated it with rotational speed as a measure
of the relative velocity Vrel
(4.73)
This result is in agreement with similar situations under turbulent conditions.
External transport limitation in this biofilm reactor was eliminated when the
linear fluid velocity exceeded 0.8 m' s -1.
Turbulence in the reactor is thought to be the parameter of greatest concern,
because liquid film resistances (kL2 and k Ll ), active microbial volume, and
mixing (te, t m) are all affected by velocity. Active microbial volume is impacted
because at higher velocities greater depth of substrate penetration increases
the active film volume available for reaction (e.g., Miura, 1976). An apparent
abrupt change in the physical properties of a biofilm reactor occurs at a critical
magnitude of Re. Data from Castaldi and Malina (1982) indicate that Re erit :""'
1000, which is less than the transition value of Re for the flow of water in a
tube. Below Re erit , the apparent uptake rate is primarily a surface reaction
regardless of Vrel'
In conclusion, it is obvious that the influence of external transport limitation
significantly falsifies the estimation of kinetic parameters. Special safeguards
are needed in using high turbulence STRs. The use of unconventional (perfect)
bioreactors for process kinetic analyses is preferable to use of STRs. As
discussed in Sect. 4.3, liquid-solid mass transfer coefficients, kL2' are corre-
lated by means of the empirical expressions referred to in Table 4.3 (similar
equations have been reported by Deckwer, 1980, and by Moo-Young and
Blanch, 1981). The relative importance of k L2 ' as depends mainly on the size
of the biomass particle, as illustrated in Fig. 4.34. In case of cell pellets of, for
example, 0.5 to 2 mm diameter, kL2 should be regarded as the major resistance
(k L2 as < kLl a), while with flocs with dp ' " 10 11m, k L2 as > kLl a, and OTR
limitation at the GIL interface will dominate.
4.5.2 INTERNAL TRANSPORT LIMITATIONS
The concept of Equs. 4.49 resp. 4.51 is also applicable in the case of internal
transport limitation; here the ratio of rate constants is treated as a dimension-
less number, the so-called "Thiele modulus" rp (Thiele, 1939) or the Damkoeh-
ler number of the second degree, Dan.
The solutions in all cases of internal transport have the form represented
by Equ. 4.55 (Emig and Hofmann, 1975) as a consequence of Equ. 4.60 with
boundary conditions 4.61a and 4.61b, representing the case of complete S
penetration within the film. Thus the relationship between ifJ and Dan is, in
general form,
krc n- 1
(4.74)
Deff
In this equation, d p is the diameter of the particle and Deff is the effective
diffusion coefficient, which is related by two-film theory to transport over a
layer of thickness d p with a transport coefficient in the solid phase of k s. The
physical meaning is the ratio between reaction rate at surface conditions and
diffusivity rate through the outer surface of the particle with diameter dp If ifJ
is small, the reaction rate becomes the rate-determining step, while diffusion
is the rate-determining step if ifJ is high.
The results of Equ. 4.55 are often given in the form of a graph; Fig. 4.35
shows the effect of internal transport on overall effective rate. Compared with
Fig. A.31, obviously the internal effect is much more pronounced than the
external. Another typical plot is shown in Fig. 4.36. In the range of kinetic
control, 11r = 1 (i.e., reff = rideal). With increasing Dan (smaller mass transport
coefficient, faster reactions), the effective rate of the reaction is throttled by
the transport limitation.
For true kinetic measurements of reaction rates in the S phase, the particle
diameter must be kept so small that ~ :::;; dcrit In Fig. 4.35b, the particle
diameter is the ordinate; this part of the figure is drawn in correspondence to
part a of the same figure. Measurements of the effectiveness factor, 11r' are
carried out experimentally by varying the particle size. If the average diameter
of the S-phase particles in an engineering process is ~ (cf. Fig. 4.2), then a plot
such as Fig. 4.36b can be used in determining the effective rate of the process.
Other methods for determining 11r are described in the literature (cf. Emig and
Hofmann, 1975).
The slope of the curves in Fig. 4.36 is only moderately influenced by the
reaction order and/or the geometry of the S-phase particles (plates, spheres,
or cylinders). However, the form of the curve is strongly affected by the
material transport coefficients and/or by Deff in the S phase. Exceptions are
possible where 11 > 1 due to nonisothermal behavior or complex kinetic
equations (Aris, 1975).
The theoretical approach must be modified in cases of incomplete S pene-
tration. When substrate is depleted within a biofilm layer <5s , which is less than
<5, boundary conditions must be modified for this fact of no mass transport
beyond z = <5s . Thus, instead of Equ. 4.61 b,
dC i =0 at z = <5s (4.61d)
dz
llr
t
:-
a
-voa;-;
---'~'NT
r eff , - - - - - - - - - - - - - ,
FIGURE 4.35. Plots of normalized overall
t
reaction rate rerrlrmax versus dimension- r max- .
less substrate concentration 51 Ks for !----------~------
different values of the modulus Jint in
case of internal transport limitation with
saturation-type kinetics. (From Horvath
and Engasser, 1974.) b
_d
l:::,.
FIGURE 4.36. Graphical representation of the concept of effectiveness factors: (a) The
effectiveness factor of reaction, 11" as a function of the Damkoehler number of the
second kind, Dan, or of the Thiele modulus, J (cf. Equ. 4.74). (b) The effective reaction
rate, rerr , as a function of the diameter of the particle, d. Part (b) can be used to obtain
the value of rerr for a given average diameter, J, of a population of floes with a
distribution d. The range of validity of kinetic control (rrds) and of diffusion control
(n rds ) is indicated in part a.
The mass balance differential equation in the case of a spherical shell of a

pellet yields, instead of Equ. 4.60
(4.75)
At steady state with a Monod-type reaction, Equ. 4.75 (with rmax = qs,max' x)
becomes
(4.76)
where ar is the biofilm surface and 5 is identical to Sr, the surface concentration.
Integration of this equation with appropriate boundary conditions would
give the concentration gradient ds/dz in the biofilm. Because it is nonlinear,
Equ. 4.76 cannot be solved explicitly. Numerous workers have presented
analytical solutions for cases when the Monod relation reduces to the first-
and zero-order approximations (Harremoes, 1977; Haug and McCarty, 1971).
Although not solving Equ. 4.76 for Sf' Atkinson and others (Atkinson and
Daould, 1968; Atkinson and Davies, 1974; Rittmann and McCarty, 1978;
Williamson and Chung, 1975) have presented solutions for the flux of sub-
strate into the biofilm for the general case of saturation-type kinetics. The
most convenient form of solution of this set of equations is obtained when
dimensionless terms are introduced, reducing the effective number of param-
eters from 6 to 3. Thus
(4.77)
(4.78)
dSf = 0 at 2 =>0 (4.79)

d2
Thereby, the parameter rjJ, representing the dimensionless biofilm thickness 15,
is defined as modulus (cf. Equ. 4.74)
(4.80)
The solution to Equ. 4.77 with 4.78 has been given in numerical form by
Schneider and Mitschka (1965) as the "biological rate equation," (cf. Equ.
5.239).
The functional representation of l1r developed for biofilms can then be
expressed as
l1r = 1 _ tanh k3 . d (~) (4.81)

k 3 d tanhrjJp
and
l1r
1
= rjJp -
tanh k3 . d
k 3 .d
(1
tanhrjJp - 1
) (4.82)
where rjJp is the generalized Thiele modulus as a function of rjJ and Ks/s. A
graphical representation of this biological rate equation is demonstrated in
Sect. 5.8.
In the case of the first-order reaction rf = kif Sf' the modulus becomes
(4.83)
with a solution, well known in literature (Levenspiel, 1972)

with kl.f [S-1].

In the case of a zero-order reaction (rf = k Of )' the Thiele modulus f3 has been
modified to suit the interpretation for biofilm kinetics. It is then the reciprocal
of the normal Thiele modulus for a zero-order reaction,
1
f3=-= (4.85)
rPo
In the case of fully penetrated biofilm, f3 is greater than 1, and thus
(4.86)
with kOf [g. cm - 3 . S -1]. In the case of incomplete substrate penetration (partly
penetrated biofilm), f3 is less than 1, and thus (Harremoes, 1977; La Motta,
1976b)
(4.87)
One concludes that a zero-order reaction in a biofilm either becomes a bulk
zero-order reaction or a half-order reaction-depending on the characteris-
tics of the biofilm (kOf' b, Deff ) and on the S concentration at the LIS interface.
The half-order reaction (see Equ. 4.87) with
kl/2 = 2Deff ' kOf [mg 1/2 . dm -1/2. S-l] (4.88)
(as recognized by Atkinson and Fowler, 1974) can easily be used as a formal
kinetic approach quantifying biofilm-processing behavior (see Sect. 5.8) (Har-
remoes, 1978).
The depth of penetration of substrate within the biofilm is given by
b 2 . Deff . s*S ) 1/2

eTit =
( k (4.89)
Of
Combining Equ. 4.89 with Equ. 4.87 results in

r eff = k Of ' A . berit [mg S-l] (4.90)
This equation illustrates that when there is incomplete S penetration in thick
films, the observed rate depends only on the magnitude of the depth of
penetration and not of the total thickness (La Motta, 1976b). This so-called
active depth was specified by Kornegay and Andrews (1969), and it is often
used for design purposes. The significance of internal mass transfer resistance
to the interpretation of the kinetic data of suspended growth systems will be
explained quantitatively. The biological rate equation, written as
s
ri,eff -- ri,max -
Ks- -+X
S
'n
'IT
(4.91)
usually is transformed for parameter estimation into one of the following

equations:
Lineweaver-Burk plot (cf. Fig. 4.24c)
- = -1+ _
rmax Ks 1
.- (4.92)
r eff '7r, v '7r, v S
Eadie-Hofstee plot (cf. Fig. 4.25)

reff r eff ' Ks
- - = '7r,v - - - - (4.93)
rmax S'rmax
Langmuir plot (cf. Fig. 4.26)

rmax
- =_ 1 +1-
Ks . _
(4.94)
reff S '7r, v '7r, v
The effect of internal mass transfer resistances can be analyzed using these
plots, as illustrated in Figs. 4.37 through 4.39 and Equs. 4.92 through 4.94
(Shieh, 1981). As can be seen, the linearity of all three linearized plots is
significantly distorted by internal mass transfer (quantified by Jfnt), especially
with large diameter particles. Larger Ks values are obtained, while rmax
(= qs,max' x) is not affected provided the S concentration is reasonably high.
If the S concentration is not sufficiently high, then both Ks and rmax would be
affected. Similar plots have been presented by different authors for comparable
situations: Ngian et al. (1977) for small and large flocs, Shieh (1979), and
Atkinson and Dr-Rahman (1979). Clearly, approximate linearity of experi-
0.5
reff Ks
0.5 _1--.--
1...-1--'---'-''-''--'-...........1...-1--'-->1
rmax 5
4.37 4.38
FIGURE 4.37. Effect of internal transport limitation on Monod-type kinetics demon-

strated on a Lineweaver-Burk plot (see Fig. 4.24c) with varied values of a modulus
,jJZ inl according to Shieh (1980a). The limiting case of no transport limitation is indicated
(rrds)'
FIGURE 4.38. Effect of internal mass transport limitation on Monod-type kinetics

demonstrated in an Eadie-Hofstee plot (see Fig. 4.25) with variation of modulus ,p2 inl
according to Shieh (l980a).
FIGURE 4.39. Effect of internal mass trans-

port limitation on Monod-type kinetics
demonstrated in a Langmuir plot (see Fig.
4.26) with varied values ofthe modulus 2 int
according to Shieh (1980a).
mental data in a certain region of conditions does not establish the absence
of a solid-phase diffusion limitation. Only by application of different floc sizes
can it be proved that l1r -+ 1.
The mass principles of the theory of mass transfer with simultaneous
reaction have been used to characterize the diffusional limitation of suspen-
sions of filamentous microorganisms (Reuss et at, 1980, 1982). If mycelial
broth can be considered as consisting of small spheres or filaments, the net
specific uptake rate of oxygen in the broth can be calculated using Equ. 4.91
with appropriate values for l1r (Aktinson, 1974) using the Thiele modulus rjJ
(see Equs. 4.80-4.82). If the assumptions concerning the hypothetical geom-
etry of mycelium structure are reasonable, then rjJ should be constant at given
fluid dynamics, and this has been experimentally verified. The effect of agita-
tion speed can be satisfactorily accounted for by rjJ. Applying the statistical
theory of turbulence (Kolmogoroff, 1941), it may be concluded that the mean
size of particles is controlled by turbulent shearing action proportional to the
energy dissipation e:
dp = cemax -0.25 (4.95)
where, according to Liepe et at (1971),
emax = 0.5 .e(~:y (4.96)
where dT = tank diameter

di = impeller diameter
e = average energy dissipation rate per unit mass
Equation 4.95 means that for elements large enough to correspond to eddies
of inertial subrange, the maximum stable diameter is only a function of e and
is independent of broth viscosity. Combining Equs. 4.95 and 4.96 and sub-
stituting into Equ. 4.80, it becomes evident that rjJ should be correlated with
the power input raised to exponent of ( - 0.25); this has also been verified by
experiments (Reuss et at, 1982). Equation 4.95 also provides an opportunity
to verify the theory for those experiments carried out with mycelium, in which
macrokinetic O 2 uptake rates varied with the impeller/tank diameter ratio

(Steel and Maxon, 1966). This experimental fact can be explained by Equs.
4.91,4.80, and 4.95: As already shown in Fig. 4.2, the effective reaction rate of
O2 uptake is indeed a function of power input and tank impeller diameter
ratio, indicating that smaller impellers are always more effective than larger
ones at a given power input per unit volume in this case of mycelium broth
aeration.
Finally, the practicability ofthe '1r concept will be illustrated for STRs. Here
it is known that due to the gradients in shear rates, floc size distribution will
be of consequence. Therefore, estimation of the biological rate equation
coefficient k3 is not possible in STRs-it is possible only in biofilm reactors.
Nevertheless, the '1r concept remains a useful approach because it has been
shown that a single floc size, iI;, (i.e., the mean value), is sufficient to characterize
a given distribution function (Atkinson and Ur-Rahman, 1979). The mean floc
size closely corresponds to the "surface" mean floc size ("Sauter diameter")
Injd?
d=i
p ;::I:;--nj--:d"'"j2 (4.97)
where nj is the number offlocs with diameter d j This mean diameter can then
be used in the graphical plot of the biological rate equation shown schemati-
cally in Fig. 4.36 to determine the effective rate to be expected for the experi-
mental situation. Clearly, Fig. 4.36 is analogous to the plots of'1r = f(~, s/Ks)
from Equs. 4.81 and 4.82 shown, for example, in Fig. 5.70.
The use of a surface-based effectiveness factor '1r.A instead of the volume-
base factors previously described ('1r = '1r. v) was suggested by Fujie et al.
(1979) and defined as
s: (-rs)dz -Deff(dS)
dz z=o
(4.98)
'1r.A = L'' ' (- rs)dz n'max
Here n'max [g. m -2. h -1] is the maximum mass flux through biofilm surface.
The physical model used by these authors was the same as that proposed by
Atkinson, extended to double S limitation (glucose and O 2 ). Their authors'
"case A" corresponds to no S or O 2 limitation, "case B" involves an inactive
aerobic film (S limitation), and "case C" involves O 2 limitation (partly an-
aerobic film). With a modified Thiele modulus ~m and a pseudo-first-order
approximation for the enzyme-catalyzed rate under the given surface condi-
tions, Fig. 4.40 shows the result of the calculation of '1r.A versus ~m as a function
of sdKs under various conditions. For biofilm reactor operation, '1r.A is more
convenient in the form
-r's = '1r.An'max (4.99)
and is simpler than
11 r,A ,.....,--,--,-.-rrrrrr-..,--..,--rn-rrr------,
A10 o~-r__~~~~____------~
I::"
.5
.5
1 '" f(s/K S )
NO LlMITATION
tD:!-LlMITATION .1
10 ~ ______ ~~~ ____ ~~ __ ~
-I
10
4.40 4.41
FIGURE 4.40. Graphic presentation ofthe surface-base effectiveness factor '1r.A (see Equ.
4.98) with Monod-type kinetics according to Fujie et al. (1979) and using a modified
modulus tPrn. The cases of no limitation (in S or 02) and O 2 limitation are indicated
in the graph with varied values of s/K s .
FIGURE 4.41. Possible cases of concentration profiles s versus z in LIS bioprocessing
with partially aerobic biofilms: (1) fully penetrated biofilm, (2) shallow biofilm, (3) deep
biofilm with incomplete penetration. Profiles A-C represent the situations of com-
bined external and internal mass transport limitations. The thickness of individual
films of mass transfer resistance (OL2, os) are indicated; 0S.A represents the active layer
of biomass (aerobic biofilm). Increasing turbulence at the LIS interface increases the
concentration profile, as shown in 0L2'
(4.100)
The conventional form of a plot of 11r, v versus rjJ is shown in Sect. 5.8.
More complex cases, such as cases with S- and P-inhibition kinetics, have
been solved numerically by Moo-Young and Kobayashi (1972). Recently
criteria have been developed specifically for Monod-, S-inhibited-, Teissier-,
and maintenance-type kinetics to quantify and predict diffusional control
within whole cells and cell floes (Webster, 1981). Further details concerning
the use of the effectiveness factor concept for the quantification of biological
processes will be given in Sect. 5.8, presenting simple formal kinetics in the
case of internal transport limitation.
4.5.3 COMBINED INTERNAL AND EXTERNAL TRANSPORT

LIMITATIONS
Normally the phenomena of external and internal mass transport cannot be

treated as completely separate. Theoretically, separation of these two effects
may be difficult, but experimentally it is quite possible to create conditions in
which (a) the effect of external transport is negligible (by increasing the relative
fluid velocity) or (b) the effect of internal limitation is minimal (by reducing
particle size). Generally, however, both phenomena must be taken into con-
sideration by checking the experimental conditions. If, for example, internal

limitation does strongly influence reaction rate, the plot of Km.app versus 1/vrel
proposed by Seong et at (1981) will not give a straight line even at high flow
rates.
The typical experimental situation in which both phenomena are significant
is shown in Fig. 4.41. In most aerobic biofilm systems, O 2 does not fully
penetrate into the whole biofilm (Fujie et at, 1979). Therefore it is convenient
to divide the overall biofilm system into four regions: bulk liquid, a diffusion
layer conforming to two-film mass transfer theory, ~2' an active or aerobic
biofilm bs , and an anaerobic or inactive biofilm. Increasing the relative fluid
velocity Vvel at the LIS interface would result in a deeper active biofilm, thus
indicating the combined action of both phenomena. Different concentration
profiles are obtained for different ratios of kL2 and ks (or Deff ), and they are
expressed by dimensionless numbers kL2 ' blDeff (see Fink et at, 1973). In the
steady state, the combined effects can be formulated using the concentration
change at the distance z = b as
n' = k L2 (CL.B- ~:) = Deff(~~)z=b (4.101)
where n' is the molar flux per unit surface area ofbiofilm [mol' m- 2 . S-I] and
CL,B and Cs' are the respective concentrations in the bulk liquid and at the
surface of the biofilm.
An analytical solution to this problem of combined external and internal
transport, calculating Cs* and n' with Equ. 4.101, is possible only for simple
reaction rate equations. A numerical solution using dimensionless forms can
be achieved in more difficult cases of enzyme kinetics (Lee and Tsao, 1972) or
by assumed first-order approximations (Rony, 1971). It is also possible to solve
the problem without numerically solving the differential equation by using
the results of computations that consider only internal transport. Horvath
and Engasser (1974) showed that external and internal transport can best be
identified using Eadie-Hofstee plots (cf. Figs. 4.33 and 4.38). In doing so, one
finds the value of Cs* where the transport of the reactant from the L bulk to
the catalyst equals the conversion in the particle. Thus
kL2 (CL.B- ~:) = b'l1r'r(cs*) (4.102)
After calculating the modulus rP by assuming a value for Cs*, I1r may be found
from existing I1r charts (Atkinson, 1974; Moo-Young and Kobayashi, 1972).
As the true value of Cs* must be found by an iterative procedure, Equ. 4.102
holds only for one particular value of cs*'
A more straightforward method for determining Cs* is a graphical method
that uses the existing I1r charts (Frouws et at, 1976). Defining a normalized
mass transport rate, 11~, in contrast to 11 .. which is a dimensionless reaction rate
(4.103)
this dimensionless factor 11~ can then be plotted against the Thiele modulus rp,
which includes Cs*, on the same graph as 11r versus rp. The intersection of 11r
and 11~ on both plots yields the correct relative values 11r and cs' needed. An
advantage of the graphical method is that it clearly shows to what extent the
diffusion resistance is inside or outside the particle. It often seems to be the
case that both effects are about equally important. In Equ. 4.104, the combined
effects of external and internal transport limitations are quantified, showing
the type of interaction
1
-reff = 1j(k. ) + 1jk sx = kapp's'x (4.104)
r 11r L2
where S == CL,B'
The behavior of this type of equation can be shown graphically by plotting
kapp against SL.B in a log-log diagram, Fig. 4.42 (Watanabe et aI., 1980).
In the case of external limitation (see profile c in Fig. 4.41), Cs is equal to
zero and Equ. 4.102 reduces to (cf. Equ. 2.2b)
(4.105)
which states that a bulk first-order reaction occurs at extremely low values of
which is in accordance with enzyme kinetics. If LIS mass transfer is
CL B '
negligible, that is, if
and (4.106)
then Equ. 4.104 is simplified to obtain Equs. 4.86 and 4.88 for the concen-
tration profiles a and b in Fig. 4.41. Case A is characterized as a zero-
r EFF 1 I---=:::::r=:::::::iiiiiii------~
r MAX
5.10
10
3.10
10 2
k 2
..- --,T-,::r,..:.TO...:.T_R_
Deft
4.42 4.43
FIGURE 4.42. Graphical plot of the relationship between apparent overall rate coeffi-
cient kapp and bulk substrate concentration SL according to Equ. 4.104. Three distinct
regions (A-C) correspond to the three cases in Fig. 4.41. (Reprinted with permission
from Prog. Wat. Tech., vol. 12, Watanabe et aI., copyright 1980, Pergamon Journals
Ltd.)
FIGURE 4.43. Effect of external Oz transfer rate kTr.tol RZ/Deff on effective respiration
rate r.fflrmax as a function of Damkoehler number Dan as a measure for particle
diameter dp (see Equ. 4.74 with n = 1 and kr = qo.maxIKs). (From Reuss, 1976.)
order reaction, where 150 is the oxygen penetration thickness, and case B is
characterized by a half-order reaction rate (cf. Equ. 4.88). Nitrification experi-
ments have been carried out, and they adequately verify the proposed model
(Watanabe et ai., 1980).
The case of coupled external and internal transports has also been dealt
with by Reuss (1976), who proposed a graphical solution to the problem,
assuming saturation kinetics for the respiration of pellets and defining an
overall transfer coefficient
1 1 1
--=--+-- (4.107)
k-rR.lol kLl aL kL2 as
which takes into account the external transports kLI aL and kL2as. With the
proper boundary condition (see Equ. 4.61), Equ. 4.75 was solved numerically
and plotted as in Fig. 4.43. Calculations demonstrate the extent to which reff
of O 2 utilization of pellets increases when external transport is increased. O2
limitation still remains as a function of pellet size, maximum respiration rate,
and the diffusion coefficient (Dan; see Equ. 4.74), and it can only be avoided
by increasing the external O 2 concentration (cf. profile Al in Fig. 4.41). On
the basis of these calculations, the potential influence of such process variables
as agitation or aeration can be predicted (cf. Fig. 4.2). In the case of the
immobilized glucose oxidase and catalase system, such considerations explain
the fact that the external mass transfer rate may cause a shift from glucose
limitation to O2 limitation (Reuss and Buchholz, 1979). This concept was also
applied to complex kinetics with Sand P inhibition (Wadiak and Carbonell,
1975), in which multiple steady states are possible and '1r > 1.
Another concept, similar to the approach just described, has been proposed
by many workers (Aris, 1975; Chen et al., 1980; Fink et ai., 1973; Gondo et ai.,
1974; Hamilton et ai., 1974). It designates an overall effectiveness factor ftr>
and it is a valuable approach when L-film resistance of mass transfer cannot
be neglected. Since ftr approaches '1r.V when LIS mass transport becomes
infinite, the latter can be regarded as a limiting case of the former (yamane,
1981). As an extension of the work of Kobayashi and Moo-Young (1973),
Kobayashi et ai. (1976) proposed the following approximate expression for
the solution of Equ. 4.77 with Equs. 4.78 and 4.79, containing terms for
reaction in the zero- and first-order regions of enzyme kinetics
(4.108)
where for biofilms
"" ~ 1
f~ qJp
(4.109a)
and
tanh r;6p
'1r.l=-~ (4.109b)
Using Equ. 4.108, Yamane (1981) attempted to estimate ryn giving different
expressions to both ry,.o and flr.l by incorporating the Biot number, Bi, as a
measure of external film resistance. For a biofilm configuration, thus
2
for r;6p :::;;
(1 + (2jBi
(4.1 lOa)
2
(1 + (2jBi
and
A (r;6p r;6;)-1 (4.110b)
11',1 = tanh r;6p + Bi
When Bi -> 00, the external transport limitation becomes negligible (cf. Equ.
4.110) and hence fI, approaches '1r' Comparing the magnitude of relative devia-
tion from the corresponding true numerical values for film and floc geometries,
Yamane showed that Equs. 4.81 and 4.82 can be recommended for a slab but
never for a sphere. The variation of the relative error with the Thiele modulus
is shown in Fig. 4.44. Equations 4.108 and 4.110 give quite similar degrees of
accuracy.
lI1)
% slab
A +2
-2
FIGURE 4.44. Comparison of dif- lI1)

%
ferent effectiveness factor con- sphere
cepts (1]" 1J~ and Ij) by demon- A
strating the variations in the
relative error ~I](%) with the
Thiele modulus I/J for different
geometry: slab (a) and sphere
(b). Atkinson concept: 1]" see
Equs. 4.81 and 4.82; Kobayashi
et al.: I]~, see Equ. 4.108; Ya-
mane: q, see Equ. 4.110. (From b
Yamane, 1981.)
10'2
FIGURE 4.45. Effectiveness factor of reaction '1r based on biofilm thickness b in depen-
dence of Thiele modulus I/J~ (see Equ. 4.111) as a function of the Thiele modulus I/J for
support particle size. (Kargi and Park, 1982.)
A Thiele modulus r/J" based on biofilm thickness (Rp - Rm) = (j
If
r/J" = (Rp - Rm) - r
Deff
=- r/J (Rp
Rm
- - 1) (4.111)
was introduced by Kargi and Park (1982) for dealing with fluidized bed
bioreactors. The effectiveness factor was shown to vary only with biofilm
thickness (r/J,,) and support particle size (i.e. modulus r/J) when the bed porosity
e was constant. This behavior is depicted in Fig. 4.45. It can be seen that
effectiveness increases as r/J decreases for a given value of r/J". There is an
optimum value of r/J" resulting in maximum effectiveness factor for any given
value of r/J. This optimum value of r/J" decreases with decreasing r/J. In contrast
to results of other work on fluidized bed biofilm reactors (cf. Sect. 6.7), this
theoretical analysis reveals that there is no dp oP' value, but rather there exists
an optimal biofilm thickness for a given support particle size, that is, an
optimal ratio of biofilm thickness/support particle diameter.
4.5.4 TRANSPORT ENHANCEMENT

Thus far only cases in which interaction between kinetic and transport factors
has led to a limitation of the reaction have been presented. We now take up
the case where the rate of transport is itself increased by the reaction taking
place.
The same method presented in Fig. 4.30 for solving the problem can also be
used here. The only difference is that the effectiveness factor is referred to the
transport rate and is called the enhancement factor, E (E == '1TR)' By analogy

to Equ. 4.51, E is defined in Equ. 4.52 and the solution equation, Equ. 4.56,
with boundary conditions Equ. 4.61c also applies.
Two extreme cases exist, as shown in Fig. 4.29 by concentration profiles c
(for E = 1) and e (for E > 1). As in all other cases, the relative magnitudes of
the rate constants define the kinetic and diffusion-controlled regimes. It is,
however, customary to describe the relative k values with a dimensionless
number, Ha. For GIL reactions with kL (kLl)
Ha =1
-
kL
J 2
--Dc
n+1
*" k
r
I
(4.112)
after the numerical, approximate solution of the differential Equs. 4.60 and
4.61a, c due to Reith (1968). The general accurate form of the solution is due
to Himer (1974):
-dc =
dt
a(c* - c) (~
- - ' D c*n-l . k ) (jl+C)
n + 1 r
(4.113)
where the integration constant, C, is actually a function of several variables
C= f ( Ha, c; ,HI) (4.114a)
although the Reith approximation uses

1
C=~ (4.114b)
Hal
The approximate solution is in error relative to the exact solution by only
+ 6~%;. The general solution is therefore also given in the form of a function of
Ha
E = f(Ha) (4.115a)
for example, that due to Reith (1968)
E = J1 + Ha 2 (4.115b)
The graphical representation of this solution is shown in Fig. 4.46. The region
Ha :::;; 0.3 is the transport-limited regime, where the effective process rate is
determined by physical adsorption. The reaction is slow here relative to the
rate of transport: It occurs completely in the liquid phase (bulk), and the
reaction therefore has no influence on the transport rate. In the same regime,
deviations due to different HI values ofGI L reactors (cf. Equ. 3.59) may occur,
giving the curves shown in Fig. 4.56 (for smaller HI values).
In the region Ha ;::-: 3, pure kinetic control is in effect, so that here the rate
of the process is determined by the rate of the reaction. The reaction is very
fast, and it occurs almost entirely in the L film at the GIL interface. In cases
where one is dealing with a suspension of catalytically active particles, a fall
E = ~Tr
A
I
I
I
./
-"'a
SLOW FAST
REACTIO~ REACTION
0,3 -Ha
FIGURE 4.46. Mass transport enhancement factor, E, or 'hR, as a function of the Hatta
number, Ha, or the catalyst concentration, CCo H, in the case of sulphite oxidation
with Ha = f(CCOH). Curves a-e correspond to those in Fig. 4.29. (Adapted from
Hirner, 1974; Reith, 1968.)
offin rate is possible if the particle diameter is comparable to the film thickness
b (cf. Fig. 4.29; Alper, Wichtendahl, and Deckwer, 1980; Deckwer and Alper,
1980).
Concentration profile 3 in Fig. 4.41 corresponds to the region Ha ;;::: 3. In
the intermediate region 0.3 < Ha < 3, represented by profile d in Fig. 4.29, the
reaction is fast enough so that it also runs partly at the GIL interface, and it
therefore also increases the rate of transport.
The distinctions made according to 0.3 ::;; Ha ::;; 3.0 are not arbitrary ones;
rather they are based on the realization that when one process takes place at
a rate 1/10th that of another process, one of the processes may be ignored (cf.
Equ.4.l1Sb).
There are many examples of enhanced transport among homogeneous
chemical processes (Dankwerts, 1970). There are several procedures for obtain-
ing kL a values: Co++ -catalyzed sulphite oxidation is a valuable method
(Reith, 1968); Cu++ -catalyzed sulphite oxidation has some kinetic problems
(cf. Sect. 3.3 and Moser, 1973); the Dankwerts CO 2 /NaOH method which also
exhibits some problems (Dank werts and Roberts, 1963); the glucose oxidase
method (Tsao and Kempe, 1960; Hsieh, Silver, and Mateles, 1969), and more
recent methods using anhydrase (Alper et aI., 1980b). They all function accord-
ing to the theory of simultaneous reaction and mass transport.
In the special case of Co ++ sulphite oxidation, the reaction rate is directly
dependent on the Co++ concentration, and consequently Ha = f(Co++).
This is taken into consideration in Fig. 4.46 in that the catalyst concentration
is shown as the second ordinate of the plot. The special importance of this
method (and also of the CO 2 /NaOH method) is that the values for kLl and a
can be obtained separately in a GIL reactor. This will be briefly described. The
solution in this case, relying on Eq. 4.113, is in case of oxygen consumption (0):
(dO)
dt eff
= a(o* _ 0) J_2+_.
n 1
D o*n-l . kr + kt (4. 116a)
It may be simplified for the region Ha ~ 3 to
(-dO)
dt eff
= a 0* J+- -2_ . D . 0 *"-1
n 1
. . kr == k L a 0* . Ha (4.117)
or, for the case Ha s 0.3, it may be simplified to

do
dt = kLa(o* - 0) (4.118)
In the intermediary region 0.3 < Ha < 3, Equ. 4.116a is fully valid, and it may
be rewritten in another form which 0 -> 0 to give
(dO)
dt eff
= kL . a. 0* E (4. 11 6b)
The experimental measurement of values for kLl and a, thus, is possible using
the following procedure:
1. Select a fast reaction (for example, Co++ > 10- 4 gmoljl) in which the
process rate constant is dependent on kr and a, according to Equ. 4.117.
The reaction rate constant kr may then be obtained as a function of tempera-
ture, pH, and catalyst concentration, assuming that one is working with a
so-called "perfect" or "model" reactor for GIL reactions in which the
interfacial area, a, is known and the hydrodynamics are simple and also
known. Examples of such reactors are the falling film reactor, the liquid jet
reactor (Astarita, 1967), and the thin film reactor with rotating drum (Dank-
werts and Kennedy, 1954; Moser, 1973) as shown in Fig. 3.43.
2. Run the fast reaction in the reactor whose value is to be determined. After
kr is known, use Equ. 4.117 to calculate the exchange surface area a.
3. Working then with the slow reaction (for example, Co++ S 10- 5 gmolj1) in
the reactor whose value is to be determined, calculate kL (a is known from
the previous experiment). Note, however, that physical factors such as
viscosity, ionic strength, surface tension, and the presence of small particles
all can have an additional influence on kL and a (Alper et al., 1980; Linek
and Benes, 1977).
The possible rate enhancement of GI L oxygen transfer by viable respirating
microbial cells was first studied by Tsao and his group (Tsao, 1968, 1969; Lee
and Tsao, 1972; Tsao et al., 1972) using a surface-agitated stirred vessel.
The importance of OTR enhancement occurring in bioprocessing was later

challenged by two other groups: Yagi and Yoshida (1975), in their experiments
in a sparged stirred tank fermenter, found on significant enhancement. Linek
et al. (1974) employed an equation based upon Dankwerts' model, calculated
very low enhancement factors (1.025), and also challenged Tsao's work. Fur-
thermore, Tsao et al. (1972) found an even more pronounced effect of OTR
enhancement due to respiration of viable cells or in a glucose oxidase system
and explained the difference between theory and experiment by the accumula-
tion of microorganisms in the surface region (so-called "two-zone model").
Yagi and Yoshida (1975) explained the discrepancy with Tsao's data by the
fact that with "young" bubbles, in which most of the mass transfer takes place,
the degree of accumulation of cells at the surface is likely to be low. As a first
conclusion, it is evident that two different problems are involved here. The
first question is whether OTR enhancement occurs at all in bioprocessing, and
the second question is why experimentally determined enhancement factors
are substantially greater than the theoretical values calculated from the anal-
ogy to chemical enhancement. The situation, and thus experimental verifi-
cation, is complex, and the reader is referred to the literature: Kung and Moser
(in press); Lohse et al. (1980); Merchuk (1977); Moser (1977c, 1980c); Sada et al.
(1981); Tsao (1978a); Tsao (1978b). Elucidation seems possible by including
cell distribution profiles in the GIL film and experiments with fixed kinetics
(e.g., enzymes) rather than microbial cells, which rapidly adapt to environ-
mental changes.
4.5.5 CONCLUDING REMARKS
Finally, we recall that the relative extent of each reaction phase and the relative
magnitudes of the rate constants determine the extent to which a pseudo-
homogeneous model approximation may be applied to a three-phase process.
If there are no transport limitations in or between phases (cf. Fig. 4.28) (if, for
example, a: -> at), one may ignore the differential equations pertaining to
mass transport, and the system of equations is reduced to that appropriate
for a pseudohomogeneous model.
In this connection, a special circumstance should be mentioned. Due to the
similarity in density between cells (p ~ 1.05) and water (p = 1), one cannot
rely on any great difference in relative velocity between Land S phases (Fig.
4.1). However, with film reactors an elevated metabolism has been found,
corresponding to the higher relative velocity that is possible in the case of
carrier-bound films of microbial cells (La Motta, 1976a; Moser, 1977b). The
kL2 value shows a corresponding dependence (power index 0.7) on flow
velocity (or rotation velocity). Relying on the effectiveness factor, one comes
to the conclusion that with elevated kL2 value, the Thiele module must be
smaller. There will thus be a concentration difference a: -> at at the LI S
interface, and with this, d eri ! in the S phase will be larger (6s. b in Fig. 4.28). The
effect of accelerated metabolism is, at least in part, interpreted as attributable
to the elimination of a transport-limiting factor, which apparently is not

completely eliminated with biofloc reactors (at =I- at). Similar effects have
been observed with pellets (Miura, 1976).
BIBLIOGRAPHY
Aiba, S. (1978). Biotechnol. Bioeng. Symp., 9, 269.

Alper, E., Wichtendahl, B., and Deckwer, W.D. (1980a). Chern. Eng. Sci., 35, 217, and
1264.
Alper, E., et aI., (1980b). In Moo-Young, M., et al. (eds.). Advances. in Biotechnology,
Vol. 1. Oxford: Pergamon Press, p. 511.
Aris, R. (1975). The Mathematical Theory of Diffusion and Reaction in Permeable
Catalysts, Vol. 1. Oxford: Clarendon Press.
Astarita, G. (1967). Mass Transfer with Chemical Reaction. Amsterdam: Elsevier.
Atkinson, B., and Daould, I.S. (1968). Trans. Inst. Chern. Eng., 46, 19.
Atkinson, B., and Daould, I.S. (1970). Trans. Inst. Chern. Eng. (Lond.), 48, T245.
Atkinson, B., and Davies, U. (1972). Trans. Inst. Chern. Eng., 50, 208.
Aktinson, B., and Davies, I.J. (1974). Trans. Inst. Chern. Eng., 52, 248.
Atkinson, B., and Fowler, H.W. (1974). Adv. Biochem. Eng., 3, 221.
Atkinson, B., and Knight, A.J. (1975). Biotechnol. Bioeng., 17, 1245.
Aktinson, B., and Ur-Rahman, F. (1979). Biotechnol. Bioeng., 21, 221.
Barford, J.P., and Hall, R.I. (1979). 7th Australian Corif. Chern. Eng., 21.
Bazine, M.I. (ed.) (1982). Microbial Population Dynamics CRC press, Boca Raton,
Florida.
Bojarinow, A.I., and Kafarow, W.W. (1972). Optimierungsmethoden in der Chemischen
Technologie (trans. from Russian). Berlin: Akademie Verlag.
Bryant, J. (1977). Adv. Biochem. Eng., 5, 101.
Cassano, A.E. (1980). Chern. Eng. Educ., Winter, 14.
Castaldi, F.J., and Malina, J.F. (1982). J. Water Poll. Contr. Fed., 54, 261.
Chen, K.-C., et al (1980). J. Ferment. Technol., 58,439.
Danckwerts, P.V. (1970). Gas Liquid Reactions. New York: McGraw-Hill.
Danckwerts, P.V., and Kennedy, A.M. (1954). Trans. Inst. Chern. Eng., 32, 51.
Danckwerts, P.V., and Roberts, D. (1963). Chern. Eng. Sci., 18,63.
Deckwer, W.D. (1980). Adv. Biotechnol., 1,203.
Deckwer, W.D., and Alper, E. (1980). Chern. Ing. Techn., 52,219.
Eadie, G.S. (1942). J. BioI. Chern., 146,85.
Eisenthal, R., and Cornish-Bowden, A. (1974). Biochem. J., 139,715.
Emig, G., and Hofmann H. (1975). Chern. Ing. Techn.,47, 889.
Environmental Protection Agency (1977). Report No. 6254-77-003a. Washington,
D.C.: u.s. Government Printing Office.
Esener, A.A., et al. (1983). Biotechnol. Bioeng., 25, 2803.
Fan, L.-T., and Wang, Ch.-S. (1968). The Discrete Maximum Principle-A Study of
Multistage Systems Optimization. New York: John Wiley.
Fink, D.J., et al. (1973). Biotechnol. Bioeng., 15,879.
Fitzer, E., and Fritz, W. (1975). Technische Chemie. Berlin-Heidelberg-New York:
Springer-Verlag.
Froment, G.F. (1975). AIChEJ, 21, 1041. Americ. Inst. Chern. Eng. Journal
Frossling, N. (1938). Beitr. Geophys., 52, 170.
Frouws, M.J., et al. (1976). Biotechnol. Bioeng., 18,53.
Fujie, K., et al. (1979). J. Ferment. Technol., 57, 99.
Gates, W.E., and Marlar, J.T. (1968). J. Water Poll. Contr. Fed., 40, R469.
Gondo, S., et al. (1974). J. Ferment. Technol., 17,423.
Hamilton, B.K., et al. (1974). AIChEJ, 20, 503. Americ. Inst. Chern. Eng. Journal
Harder, A., and Roels, J.A. (1982). Adv. Biochem. Eng., 22, 56.
Harremoes, P. (1977). Vatten, 2, 122.
Harremoes, P. (1978). Water Poll. Microbiol., 2, 71.
Hartmeier, W. (1972). Ph.D. Thesis, Technical University, West Berlin.
Haug, R.T., and McCarty, P.L. (February 1971). Techn. Rep. No. 149, Civil Engineering
Department, Stanford University, Stanford, Calif.
Hirner, H. (1974). Ph.D. Thesis, Technical University, Stuttgart.
Hofmann, H. (1975). Chimia, 29, 159.
Hofstee, B.H.J. (1952). J. BioI. Chern., 199,357.
Horvath, e., and Engasser, J.-c. (1974). Biotechnol. Bioeng., 16,909.
Howell, J.A. and Atkinson, B. (1976). Biotechnol. Bioeng., 18, 15.
Hsieh, D.P.H., Silver, R.S., and Mateles, R.I. (1969). Biotechnol. Bioeng., 11, 1.
Kargi, F., and Park, J.K. (1982). J. Chern. Tech. Biotechnol., 32, 744.
Kobayashi, T., and Moo-Young, M. (1973). Biotechnol. Bioeng., 15,47.
Kobayashi, T., et al. (1976). J. Ferment. Technol., 54, 260.
Kolmogoroff, A.N. (1941). Compt. Rend. Acad. Sci. USSR, 30, 301; and 32,16.
Kornegay, B.H., and Andrews, J.F. (1968). J. Water Poll. Contr. Fed., 40, 460.
Kornegay, B.H., and Andrews, J.F. (1969). Proceedings of the 24th Industrial Waste
Conference, Purdue, Univ. Press p. 1398. Lafayette, Ind.
Kossen, N.W.F. (1979). In Symp. Soc. Gen. Microbiol., 20, 327.
Kling, W., and Moser, A. Biotechnol. Lett. (in press). 8.
La Motta, E.J. (1976a). Biotechnol. Bioeng., 18, 1359.
La Motta, E.J. (1976b). Environ. Sci. Technol., 10,765.
Langmuir, I. (1918). J. Am. Chern. Soc., 40, 1361.
Lee, S.B., and Ryu, D.D.K. (1979). Biotechnol. Bioeng., 21,1499.
Lee, Y.H., and Tsao, G.T. (1972). Chern. Eng. Sci., 27, 2601.
Liepe, F. et al. (1971). Chern. Tech., 23, 23. Chemische Technik
Lilly, M.D., et al. (1966). Biochem. J., 100,718.
Linek, V., and Benes, P. (1977). Biotechnol. Bioeng., 19,565.
Linek, V., and Turdik, J. (1971). Biotechnol. Bioeng., 13,353.
Linek, V., et al. (1974). Chern. Eng. Sci., 29, 637.
Lineweaver, H., and Burk, D. (1934). J. Am. Chern. Soc., 56, 658.
Lohse, M., et al. (1980). Chern. Ing. Techn., 52, 536.
Luft, G., and Herbertz, H.A. (1969). Chern. Ing. Techn., 41, 667.
Merchuk, J.e. (1977). Biotechnol. Bioeng., 19, 1885.
Meyrath, J., and Bayer, K. (1973). In Proc. 3rd Symposium Technische Microbiologie
in Berlin, Dellweg, H., ed., Inst. f. Giirungsgewerbe und Biotechnolgie, p. 117.
Miura, Y. (1976). Adv. Biochem. Eng., 4, 3.
Monod, J. (1942). Recherches sur la Croissance des Cultures Bacteriennes. Paris:
Hermann.
Moo-Young, M., and Blanch, H.W. (1981). Adv. Biochem. Eng., 19, 1.
Moo-Young, M., and Kobayashi, T. (1972). Canad. J. Chem. Eng., 50,162.
Moser, A. (1973). Chem. Ing. Techn., 45, 1313.
Moser, A. (1977a). Thesis, Habilitation, Technical University, Graz, Austria.
Moser, A. (1977b). Chimia, 31,22.
Moser, A. (1977c). Chem. Ing. Techn., 49, 612.
Moser, A. (1978). 1st European Congress on Biotechnology, Interlaken, Switzerland,
Part 1 p. 88. Dechema, Frankfurt.
Moser, A. (1979). Paper presented at Training course "Genie chimique dans les Opera-
tions Biologiques". Societe de Chimie Industrielle, Bruxelles, November 26-28.
(Prof. R. Jottrand, ed.) Univ. Libre de Bruxelles.
Moser, A. (1980a). In Proc. UNEPjUNESCO/ICRO Course, Theoretical Basis of
Kinetics of Growth, Metabolism and Product Formation of Microorganisms Jena,
East Germany: Academy of Sciences, Zentralinstitut fUr Mikrobiologie und Experi-
mentelle Therapie, Part II, p. 27. (Knorre W., ed.)
Moser, A. (1980b). In Moo-Young, M., and Robinson, C.M. (eds.). Waste Treatment and
Utilization, Vol. 2, p. 177.
Moser, A. (1980c). In Ghose, T.K. (ed.) Proceedings of the 2nd International Symposium
on Bioconversion and Biochemical Engineering. Vol. 2. New Delhi: lIT, p. 253.
Moser, A. (1981). Bioprozesstechnik. Vienna-New York: Springer-Verlag.
Moser, A. (1982). In Proceedings of the 3rd Austrian-Italian- Yugoslavian Chemical
Engineering Conference, Graz, (Moser F., ed.) T.U. Graz, Austria, September 14-16,
Vol. 2, p. 620.
Moser, A. (1983). Proc. Adv. Ferm. 83 (Suppl. Process Biochemistry), 221.
Moser, A. (1984). Acta Biotechnolog., 4, 3.
Moser, A., and Lafferty, R.M. (1976). In Dellweg H. (ed.) Proc. 5th International
Fermentation Symposium, Berlin, Abstract No. 5.21.
Ngian, K.F., et al. (1977). Biotechnol. Bioeng., 19, 1773.
Oosterhuis, N.M.G. (1984). Ph.D. Thesis, Technical University, Delft, Netherlands.
Pickett, A.M., and Topiwala H.H., and Bazin, M.J. (1979). Proc. Biochem., 14, 10.
Pitcher, W.H. (1978). Adv. Biochem. Eng., 10, 1.
Prigogine, I., and Defay, R. (1954). Chemical Thermodynamics. London: Longman
Green.
Reith, T. (1968). Ph.D. Thesis, Technical University, Delft, Netherlands.
Reuss, M. (1976). Fort. Verfahrenstechnik,551.
Reuss, M., and Buchholz, K. (1979). Biotechnol. Bioeng, 21, 2061.
Reuss, M., and Wagner, F. (1973). In Dellweg, H. (ed.). Proceedings of the 3rd Sym-
posium on Technical Microbiology. Inst. fUr Garungsgewerbe und Biotechnologie
Berlin:, p. 89.
Reuss, M., et al. (1980). Paper presented at 6th International Fermentation Symposium,
London, Ontario.
Reuss, M., et al. (1982). J. Chem. Tech. Biotechnol., 32, 81.
Ringpfeil, M. (1980). Paper presented at 2nd Symposium on Socialist Countries on
Biotechnology. Leipzig, East Germany 2-5 December.
Rittmann, B.E., and McCarty, P.L. (1978). J. Environ. Eng. Div., ASCE, 104 (EE 5), 889.
Rittmann, B.E., and McCarty, P.L. (1980). Biotechnol. Bioeng., 22, 2343, 2359.
Rittmann, B.E. (1982a). Biotechnol. Bioeng., 24, 501.
Rittmann, B.E. (1982b). Biotechnol. Bioeng., 24,1341.
Roels, J.A. (1982). J. Chern. Tech. Biotechnol., 32, 59.

Roels, J.A. (1983). Energetics and Kinetics in Biotechnology. Amsterdam: Elsevier
Biomedical Press.
Romanovsky, J.M., Stepan ova, N.V., and Chernavsky, D.S. (1974). Kinetische Modelle
in der Biophysik, Jena: VEB G. Fischer.
Rony, P.R. (1971). J. Chern. Tech. Biotechnol., 13,431.
Sada, E., et al. (1981). Biotechnol. Bioeng., 23,1037.
Sano, Y., Yamaguchi, N., and Adachi, T. (1974). J. Chern. Eng. (Jap.), 7(4), 225.
Schreier, K. (1975). Cherniker Zeit., 99,328.
Schneider, P., and Mitschka, P. (1965). Colin. Czech. Chern. Cornrnun., 30,146.
Seong, B.L., et al. (1981). Biotechnol. Lett., 3, 607.
Shah, Y.T. (1979). Gas-Liquid-Solid Reactor Design. New York: McGraw-Hili.
Shieh, W.K. (1979). Biotechnol. Bioeng., 21, 503.
Shieh, W.K. (1980a). Water Res., 14,695.
Shieh, W.K. (1980b). Biotechnol. Bioeng., 22, 667.
Shieh, W.K., et al. (198l). J. Water Poll. Contr. Fed., 53, 1574.
Steel, R., and Maxon, W.D. (1966). Biotechnol. Bioeng., 8, 97.
Sweere, A., et al (1987). Enzyme and Microbial Technology, 9, 386.
Sylvester, N.D., Kulkarni, A., and Carberry, J.J. (1975). Canad. J. Chern. Eng., 53, 313.
Thiele, E.W. (1939). Ind. Eng. Chern., 31, 916.
Toda, K. (1975). Biotechnol. Bioeng., 17, 1729.
Toda, K. (1981). J. Chern. Tech. Biotechnol., 31, 775.
Tsao, G.T., and Kempe, L.L. (1960). Biotechnol. Bioeng., 2,129.
Tsao, G.T. (1968). Biotechnol. Bioeng., 10,765.
Tsao, G.T. (1969). Biotechnol. Bioeng., 11, 1071.
Tsao, G.T. (1978a). Biotechnol. Bioeng., 20, 157.
Tsao, G.T. (1978b). Chern. Eng. Sci., 33, 627.
Tsao, G.T., et al. (1972). In Proceedings of the 4th International Fermentation Sym-
posium Kyoto, Japan, p. 65.
Tsao, G.T., et al. (1977). Biotechnol. Bioeng., 19,557.
Wadiak, D.T., and Carbonell, R.G. (1975). Biotechnol. Bioeng., 17, 1761.
Walker, A.C., and Schmidt, C.L.A. (1944). Arch. Biochern., 5,445.
Watanabe, Y., et al. (1980). Prog. Water Tech., 12,233.
Webster, LA. (1981). J. Chern. Tech. Biotechnol., 31,178.
Wilderer, P. (1976). In Hartmann, L. (ed.). Karlsruher Berichte zur Ingenieurbiologie,
Vol. 8, Karlsruhe, West Germany: University of Karlsruhe.
Williams, F.M. (1967). J. Theoret. Bioi., 15, 190.
Williams, F.M. (1975). In Pattern, B.V. (ed.). System Analysis and Simulation in Ecology,
Vol. 1. New York: Academic Press, Chap. 3, p. 197.
Williamson, KJ., and Chung, T.H. (1975). Paper presented at 49th National Meeting
of the American Institute of Chemical Engineers, Houston.
Wuhrmann, K. (1963). In Eckenfelder, W.W., and MacCabe, J. (eds.). Advances in
Biological Waste Treatment. New York: Pergamon Press, p. 27.
Yagi, H., and Yoshida, F. (1975). Biotechnol. Bioeng., 17, 1083.
Yamane, T. (1981). J. Ferment. Technol., 59, 375.
Yoshida, F., and Yagi, H. (1977). Biotechnol. Bioeng., 19,561.
CHAPTER 5
Bioprocess Kinetics
Model building is considered here as an adaptive process (cf. Fig. 2.18): It

involves stepwise fitting of the parameters, k, and discrimination of the
function, j, itself. All of the models in this chapter should be considered as
working hypotheses. The nature of formal kinetic descriptions means that
other mathematical functions can always be found that serve the same descrip-
tive function to within the precision of the measurements (see Esener et al.,
1983). The lack of basic content in formal kinetic models in comparison with
structured models (Harder and Roels, 1981; Roels and Kossen, 1978) can to
some extent be compensated for by subsequent analysis (Esener et al., 1983;
A. Moser, 1978b, 1984a).
The basic kinetic scheme for a typical bioprocess is presented in Fig. 5.1 as
the starting point for the analysis (A. Moser, 1980a). A chosen rate, somehow
connected with the biological growth process, is plotted as a function of
a chosen, relevant quantity-for example, a concentration, the pH, or the
temperature. One observes a typical pattern offour regions: maintenance level
endogenous metabolism, stimulation, inhibition, and toxicity. The typical
biological optimum, with r max intermediate between two critical concentration
values, is a consequence of opposing stimulatory and inhibitory factors. The
influence of a transport limitation is schematically illustrated by the quantity
representing an effective diffusion coefficient Deff in the liquid and/or solid
phase. The goal of a formal kinetic analysis is description of the various factors
influencing this course of events.
To a large extent, the formal kinetic analysis techniques presented in this
chapter relate to discontinuous batch operations. Even if the goal is a con-
tinuous operation, the batch process kinetic model serves as a start-up. The
most significant element of a kinetic analysis is the time dependence of the
macroscopic process variables mentioned in Chap. 2. Bacteria, molds, viruses,
and yeasts all have different reproduction mechanisms, and formulating a
structured kinetic model more closely related to the actual mechanism is
a desirable goal. More structured models are desirable not only to deal with
active cells but also to extend kinetic analysis to more complex situations
involving inactive cells, mixed populations of cells, multiple substrates, and
198 5. Bioprocess Kinetics
REACTION RATE

BIOLOGICAL
OPTIMUM
rmax ---- - SATURATION TYPE KINETICS
o i--=:"'-~'----:--:----~-- PROCESS VARIABLE E.G.:

CRITlCAL VALUES CON CENTRATI ONS
TEMPERATURE
1 2
e
I II
~-.~----.--.-~- --.-~.~
III
.. IV
la ENDOGENOUS STIMULATION INHIBITION TOXICITY

r'lET ABOLI SM
AND/OR
IbBIOLOGICAL
INERTIA
FIGURE 5.1. Schematic of the general kinetic behavior of bioprocesses in terms of the
various regions ofthe effect of reaction (growth or substrate utilization) rate influencing
factors. (From A. Moser, 1980a.)
so on. However, with a formal approach dealing mainly with experimental

observations, such structuring is undertaken not for its own sake: Rather,
structuring primarily serves the objective of improving the accuracy or ex-
tending the range ofthe model. Computer simulations also play an important
role in formal kinetic analysis (cf. App. II).
5.1 Temperature Dependence, k(T), Water Activity, aw,

and Enthalpy jEntropy Compensation
In Equ. 2.53, the very general model equation r = f(x, k) is shown separated
into concentration- and temperature-dependent parts. The reaction rate con-
stants are therefore regarded as temperature dependent, as illustrated with
type 3 in Fig. 4.12.
For a general equation, especially one for use in food technology (e.g.,
drying processes), the water concentration or water activity, aw , should be
incorporated. The water activity is defined as
5.1 Temperature Dependence and Enthalpy/Entropy Compensation 199
FIGURE 5.2. Dependence of the relative rates of various processes on water activity
at constant temperature. Curve 1: relative destruction rate of chlorophyll in spinach
at 37C. Curve 2: relative decay rate of ascorbic acid. Curve 3: relative oxidation rate
of potato chips stored at 37C. Curve 4: relative rate of inactivation of phosphatase in
skim milk. Curve 5: relative rate of inactivation of Clostridium botulinum. Curve 6:
drying rate of a slab of glucose at 30C slab temperature. (From Thijssen, 1979.)
aw = (PH,oh.material
- (S.1)
(PH 2 0h
and can be experimentally determined using hygrometers or the freezing point
depression method (Bol et aI., 1980).
The rate of drying and of loss of aromatic components, and also the rates
of enzymatic, chemical, and physical processes including the loss of microbial
cells, are all dependent on a w (Thijssen, 1979), as shown in Fig. S.2. Thus
(S.2)
The T dependence of k generally follows the Arrhenius equation from chemical
reactions
(S.3a)
where koo and Ea are determined from the experimental dependence of aw .

The value kco is the maximum value of the rate constant, which, according
to collision theory or the theory of activated complexes, continues to contain
a T dependence. Deviations are thus found at high temperatures. Ea is the
activation energy in [kJ mole -1]. A linearization of Equ. S.3a is possible using
Equ. S.4, which represents the Arrhenius plot
(S.4a)
or
Ea 1
Ink = Inkro - -
R' -
T (S.4b)
when Ea is considered T independent. With Equ. 5Ab the parameters Ea and

koo can be estimated using Arrhenius plots.
For microbial processing one must expect a certain complexity in the law
of temperature variation, since growth depends on a sequence of reactions.
In formal analogy to chemical reactions, an approximate agreement with
the Arrhenius law can be considered a suitable working hypothesis, because
the effective growth rate is the rate of synthesis minus the rate of degeneration.
Thus
k.
bID
= k 00, 1 e-Ea,,fRT - k 00, 2 e- E',2/RT (5.5)
where E a ,2 will be considerably greater than Ea ,l (Hinshelwood, 1946).
At low values of T, the second term is negligible, and the rate will formally
follow the simple Arrhenius law with E a , l ' Over a certain narrow range of T,
the two terms will be of the same order of magnitude. After that the negative
second term far outweighs the first; the rate will fall rapidly to zero with
a formal activation energy of microbial death E a,2'
This plot is shown in Fig. 5.3 for a microbiological process (Humphrey,
1978). Corresponding to the dual character of T, which first stimulates growth
and then, when high, destroys the microorganisms, the Arrhenius plot has
two straight lines. The slopes of these lines correspond to the activation
energies of growth, Egn identical with E a , l' and of the killing process, Ed'
identical with Ea 2
The rate constants in the case of biological processes are, for example, J.1max
and kd . The parameter Ks is also T dependent, as may be seen from the
definition in App. III.
Ink
FIGURE 5.3. Kinetic effect of temperature on a biocatalytic process: Arrhenius plot of

the growth and death of microbial cells. The apparent activation energies Eal for
growth (Egr) and Eaz for death (Ed) can be estimated from the slopes, and the factor
kco from the intercept.
Ink
Ink
a ~ ______________ 1 ~ _____________ 1 b
T T
FIGURE 5.4. Deviations from simple Arrhenius equation kinetics of bioprocesses. (a)
Sudden change in activation energy (El ~ E 2 ) due to a second enzyme that becomes
rate limiting. (b) Continuous change of activation energy to another value (after Talsky,
1971). Quantification is possible by separating the problem into segments in which
separate linear approximations (El' E 2 , E') are valid.
It is an experimental fact that the Arrhenius equation is applicable to

bioprocesses. A theoretical prerequisite for the Arrhenius equation is well
known-that the velocities of, for example, gas particles follow a Boltzmann
distribution. The distribution becomes narrower with increasing molecular
weight, so that in the case of enzymes and cells it barely exists. Without the
theoretical background, application of the Arrhenius equation to a micro-
biological processes must be considered a formal kinetic procedure with
appropriate fitting parameters.
Deviations from the simple Arrhenius relationship are known: They may,
in principle, correspond to Fig. 5.4. Part (a) of this figure represents the sudden
change from one rate-determining step to another; for example, from enzyme
E 1 , with its activation energy, to enzyme E 2 , and its own activation energy.
In part (b) the transition is more gradual (Tal sky, 1971).
The following functions are useful for linear, exponential, and hyperbolic
cases as alternatives (see also Equ. 2.57):
J(T) = ex + f3. T (5.6a)
J(T) = ex TP (5.6b)
J(T) = f3: T (5.6c)
where ex and f3 are empirical coefficients (Saguy and Karel, 1980). In situations
where the Arrhenius equation is not applicable, the best correlation is usually
found with hyperbolic equations. Another numerical fit to the typical optimum
curve, as schematically represented in Fig. 5.1, is achieved with Equ. 5.7
J(T) = ex(k - T)P. ey(k-T) (5.7)
used by Setzermann et al. (1982).
In reff In N
_ n RDS :
_._
INTERNAL/EXTER"iAL :
TRANSPORT LIMITATION'
a
1
L-------------------~--T
6.
FIGURE 5.5. Alteration of the observed
activation energy due to the influence of
mass transport (after Dialer and Lowe,
1975).
FIGURE 5.6. Alternative concepts for [> D2
quantification of temperature effects: (a)

b
"decimal reduction time," D 10 (see Equ.
~--~~------~~---------T
5.8), and (b) "Z value" (see Equ. 5.9).
Further deviations are the result of transport influences represented in

Fig. 5.5 by an Arrhenius-type plot for a case with strong internal and external
diffusion control (Dialer and Lowe, 1975). One can see that only for the lowest
temperatures is the value of E calculated from the linear slope a true value.
An additional alternative to the Arrhenius law for correlation of T depen-
dencies is given by the concept of "decimal reduction time," or the "D 10 value,"
which is used in the food-processing industry. DlO is defined as the time
required to reduce a microbial population by one 10glD cycle, as illustrated in
Fig. 5.6a. The T dependence of the Dl 0 value is characterized by the "Z value,"
as shown in Fig. 5.6b. The following relations apply:
2.3
D 10 = - (5.8)
kd
Tz - Tl
Z = .,------------ (5.9)
log D z - log Dl
The D1D concept is adequate only over very narrow T intervals, because
according to the Arrhenius equation, Fig. 5.6b is linear only in a narrow range.
Difficulties in analyzing the T dependence are given by the facts that
- The effects of T and pH are, in general, not independent. Only pH-corrected
parameters should be used.
- Very little significance can be attached to studies on the T dependence

of enzyme kinetic parameters, unless the mechanistic meaning of these
parameters is known.
- Comparison of the activation parameters ft..H and ft..S is often much more
informative than comparison of simple rate constants.
Despite difficulties of interpretation, the application of theoretical concepts
permits conclusions, or at least hints at conclusions, concerning the mechanism.
As a result of the application of the theory of activated complexes, Equ. 5.3a
can be recast as
k (k\ T.
= e-Ll.S"/R). eLl.H"/RT (5.3b)
where ft..H # = enthalpy of formation of activated complex [J . mole -1 ]

ft..S# = entropy [J. K -1. mole- 1 ]
kB = Boltzmann constant [1.38 x 10- 16 ergtC]
h = Planck constant [6.6 x 10- 27 erg sec]
The expression within the brackets of Equ. 5.3b is identical with the factor
kw Taking the derivative of Equ. 5.3b one obtains (Johnson, Eyring, and
Polissar, 1954):
k kB ft..S# ft..H# 1
In - = In - + - - - -_.- (5.10)
T h R R T
From the slope of the straight line in the Arrhenius plot (In ~ vs. ~) one
obtains ft..H#, and from the intercept one may calculate ft..S#. For a particular
reaction in an aqueous medium, if one then plots ft..H # against ft..S# for various
conditions, one often finds a linear relationship, as illustrated in Fig. 5.7.
FIGURE 5.7. Enthalpy/entropy plot (I1H" /115") ac-

cording to Equ. 5.11 exhibiting a linear relationship
("linear enthalpy/entropy compensation law"). From
the slope of the line the isokinetic temperature Tp can
be estimated. -115*
The slope of this straight line is referred to as the "isokinetic temperature,

Tp" (Barnes, Vogel, and Gordon, 1969; Leffler, 1966). This relationship with
a linear thermodynamic compensation of rates is called "enthalpy/entropy
compensation"; it is important for the physiological stability of proteins
(Lumry and Eyring, 1954). The value Tp is generally between 2700K and
320 o K; for the thermal destruction of cells Tp is between 320 K and 350 o K.
0
The importance of this fact is the possible implication of a uniform mechanism

for cell death through protein denaturation.
In considering enthalpy/entropy compensation, using the basic laws of
thermodynamics
(5.11)
Equ. 5.10 may be rewritten. Microbial cell death, for example, can be related
to the optimal growth temperature of yeast cells, Tmax A slightly modified
form of Equ. 5.11
(5.12)
with
I1C{max+n = c(Tmax + n) (5.13)
can successfully be used for this purpose where c and n are empirical coefficients
(van Uden, Abranches, and Cabeca-Silva, 1968).
Even though, in many cases, compensation may be considered artifactual
(van Uden and Vidal-Leiria, 1976), this method is nevertheless a useful
instrument for predicting rate constants and activation energies under various
conditions.
The variation of Ea with respect to aw can be explained in this way, without
resorting to a particular change in mechanism (Labuza, 1980).
5.2 Microkinetic Equations Derived from the Kinetics

of Chemical and Enzymatic Reactions
5.2.1 THE DYNAMIC FLOW EQUILIBRIUM ApPROACH TO
LIFE PROCESSES
The so-called "dynamic flow equilibrium" is a basic consideration in the

kinetics of biological systems following the concept of von Bertalanffy (1942)
(see also Netter, 1969, and von Bertalanffy et aI., 1977). The dynamic flow
equilibrium takes into account once again the conservation of mass (cf.
Equ. 2.3a). In open systems, with continuous input and output, reversible
reactions take place, but the entire process may be irreversible because of
transport. Even for the simplest stationary-state case, the reaction equation,
which is
A ---+ A (k+I) B ---+ B (5.14)
ex DA k-l DB ex
5.2 Microkinetic Equations Derived from the Kinetics 205
with the transport of external materials A and B (diffusion constants DA

and DB and transport coefficient kTR ), the typical dynamic flow equilibrium
behaviors are manifest, such as:
1. The steady-state concentrations, cj , are constant at constant flux.
2. A time-dependent or permanent change in the dynamic equilibrium can
come about only through a change in the kinetic parameters k 1 (that is,
through enzyme concentration changes). From a second simple example
k+l
A ex ---+
DA
A E
k-l
I B
!k 2 (5.15)
c ---+
Dc
Cex
further characteristics of the dynamic equilibrium of life processes can be
deduced.
3. The conversion ratio of the reactants is constant and independent of external
concentrations, which means that the reaction is self-directing (autoregula-
tion, or "equifinality"). This fact becomes clear from the following type of
equation, which is the steady-state solution of the system in Equ. 5.15:
- k+1 k2
a:b:c= 1 : - : - - (5.16)
Ll krR.C
with
1
a= a ex (5.17)
1 + k 2/krR.A
Equation 5.17 shows the typical result of all open systems (living cells,
continuous flow reactors), that the ratio of rate coefficients of reaction to
transport determines the system behavior (cf. Equ. 4.62c).
All open systems, not just living ones, seem to operate as if they had a goal
of maintaining their dynamic equilibrium. This has the effect of making the
system independent of the environment. For example, in a negative feedback
system, a temporary disturbance in equilibrium constants will cause the
system to oscillate until the old steady state is reestablished. If the disturbance
is permanent, a new steady state will be found.
A well-known case is the glycolytic pathway, where oscillations in glycolysis
and respiration are the result of feedback activation and inhibition of the
allosteric oscillophores phosphofructokinase (pfk) and pyruvate kinase (pyk),
which are stoichiometrically coupled to each other through ATP/ADP and
AMP, or through NADHjNAD in case of glyceraldehyde 3-phosphate
dehydrogenase (gapdh) and alcohol dehydrogenase (adh). This system is sche-
matically represented in Fig. 5.8, a block diagram of glycolysis that has been
used in computer simulations (e.g., Garfinkel, 1969; Heinzle et aI., 1982; Hess
and Boiteux, 1967; Higgins, 1967; Mochan and Pye, 1973). In the network
sequences of metabolic pathways regulatory enzymes underlie the allosteric
effect (Hill kinetics), especially when situated at the beginning and/or end of
reaction sequences.
HEXOSE
HK
FIGURE 5.8. Block diagram of glycolysis in yeast metabolism including stoichiometric

(- and informative (- - -.) interactions. (From Hess and Boiteux, 1973.)
5.2.2 CONTRIBUTION OF ENZYME MECHANISM TO BIOPROCESS

KINETIC MODELS
5.2.2.1 Simple Enzyme Reactions

Microbial activities like growth and product formations can be regarded as
a sequence of enzymatic reactions. On this basis Perret (1960) constructed
a kinetic model for a growing bacterial cell population. The main pathways
for major nutrients are considered together with pathways for minor nutrients
and trace elements linked to each other. This metabolic network can be
simplified with the aid of the concept of the rate-determining step (rds),
resulting in a "master reaction" or bottleneck that limits the total flux and
the rate of the process.
This picture can be used to demonstrate that enzymatic reactions are the
center of kinetic considerations for microbial systems, even if it is hard to verify
directly such a mechanistic approach in real fermentations.
The differences between steady state, stationary state, and transient state
may be clarified with the aid of a consecutive reaction of order n = 1:
(S.18)
with the differential equations
da
(S.l9a)
dt
db
-dt='
kr 1 a - kr . Z b (S.l9b)
de
-dt = kr . z'b (S.19c)
The concentration time curve is presented in Fig. S.9, which shows the ranges
of the different states mentioned.
The following equation summarizes all possible reaction mechanisms con-
taining the reversible steps of adsorption, desorption, and reaction:
(S.20)
From this generalized mechanism, all singular kinetic equations known in the
literature can be derived (see App. III), including the following five types of
kinetics:
FIGURE 5.9. Illustration of different states in reaction kinetics in case of a simple

consecutive chemical reaction with zero order in the first step (Ok!) and first order in
the second ek 2 ): transient states, steady state, and stationary state.
1. Michaelis- Menten equation (cf. Equ. 2.54) in case of irreversible reaction

(L z = 0) and of no distinction between ES and EP (k+3 = 0 = L3).
2. Briggs-Haldane equation as an alternative with a different interpretation
of the Km value.
3. Langmuir- Hinshelwood approach, with irreversible reaction as further
alternative applied to chemical and enzymatic reactions (L z = 0).
4. The reversible case with L2 > 0 for Michaelis-Menten kinetics (k+3 =
0= L3).
The resulting final kinetic equation (e.g., Mahler and Cordes, 1966)
rtot _- 1.+ -
. _ r+maxr-maxS - r+maxr-maxpIKeq
1_ - ----------.......:...""---~ (5.21)
Ks r -max + r -max S + r+ max PIKeq
is valid not only for the so-called "initial rate kinetics in the forward
direction" (cf. Equ. 2.54), as it is utilized in biochemistry, but over the entire
range, induding the final phase where the reverse reaction (r _) is noticeable,
until the equilibrium between r+ and r_ is established. The equilibrium
constant Keq also called the "Haldane relationship," shows the limitation
of reversible enzymatic reactions due to the thermodynamic equilibrium:
K = k+ 1 k+ 2 (5.22)
eq k-l . k-z
The individual reaction rates (r + at P = 0 and r _ at S = 0) govern simple
enzyme kinetics. The kinetic parameters, among others, are taken from
Equ. 5.22 using special methods for experimental verification.
5. The reversible Langmuir- Hinshelwood kinetic equation, utilizing the full
mechanism in Equ. 5.20, results in the following equation:
r +max . Kp . S - r -max Ks . P
rtot = r_ - r_ = ---~------- (5.23)
KsK p + Kps + Ksp
and doubtless it presents a realistic representation useful for engineering
calculations in situations in which the accumulation of product inhibits
further production due to the reversible nature of the reaction (cf. App. III).
The evaluation of the parameters of the model is complex in all cases of

reversible reactions. For various P concentrations, one draws, for example,
a double reciprocal plot (similar to Fig. 4.24c) of llr versus lis; all of the
straight lines pass through the point 1/rmax. The Ks value is greater with greater
P concentration. The kinetic parameters can be calculated with the help of
the initial rate parameters r+ and r_.
In the same spirit as the simple reaction mechanisms already discussed
(cf. Equ. 5.20 and App. III), and their final kinetic equations (Equs. 2.54, 2.55,
5.21, 5.23), mechanisms and kinetic equations can also be derived for the
inhibition of various types of enzymatic reactions. A special case has already
been considered in Equ. 5.23. The most important class of inhibitors reducing
the rate of enzyme-catalyzed reactions is that of reversible inhibitors forming

dynamic complexes with an enzyme whose catalytic properties differ from
those of the free enzyme. A general scheme proposed by Botts and Morales
(1953) is useful in discussing inhibitors that are not reaction products:
]r
S
E' ......... ES
1 ~[~: I (5.24)
EI """'- EIS
'7
S
where I is the inhibitor. This scheme includes most of the simple types of
inhibition:
1. Competitive inhibition if EIS and the reactions involving it are missing
2. Uncompetitive inhibition if EI is missing and EIS does not break down
3. Noncompetitive inhibition if EIS complex is inactive and no effect on
substrate binding occurs (E and EI have equal affinities for S)
The kinetic behavior of various inhibition types is represented in the graphs
of Fig. 5.10a-c. An additional type of inhibition is the so-called "substrate
inhibition," which can occur at high S concentrations, graphically shown in
Fig. 5.10d (cf. Equ. 5.88). The resulting kinetic equations are:
1/r
a c
----LL-~--------------1/s ~~~~~~------------1/s
-III)"
b
d
--..L.----....L.-----------1/s -------=-.L....-------------1/s
-III)" -I/Km
FIGURE 5.10. Demonstration of the four basic types of inhibitions of enzyme kinetics
in Lineweaver-Burk plots (see Fig. 4.24c): (a) competitive, (b) noncompetitive, (c)
uncompetitive, and (d) substrate inhibition. The parameters rmax and Km can be
estimated from the intercepts and slope of the line with p = 0, where p = inhibitor
concentration.
For noncompetitive inhibition
r= (S.2S)
1 + ilK, Ks +S
For competitive inhibition
S
r=r (S.26)
max [Ks/(l + ilK,)] + s
For uncompetitive inhibition
rmax
r= . - - - -S - - - - (S.27)
1 + ilK, [Ks/(l + ilKdJ + s
A comparison of these equations shows that the various types of inhibitors
(with concentrations i and inhibition constants Kd result in a change of rmax
or Ks or both of these parameters. An approximate, graphical presentation
can show this (cf. Figs. S.lOa-c and S.11) with p = i.
One can use a plot of llr versus Iii (the Dixon plot) to obtain the parameter
Kb as demonstrated in Fig. S.l2a-c. The Walker plot can also be used for this
purpose (Wilderer, 1976) as an integral evaluation method (cf. Fig. 4.20).
Although there are basic types of inhibition (see preceding section), in
practice inhibition has mixed effects on multi substrate enzymes and multiple
inhibitors, and further classification is necessary:
1. Linear, if the replot of slope andlor intercept versus i is a straight line
2. Parabolic, if the slope andlor intercept replot is a parabola
3. Hyperbolic, if the replot is a hyperbola (Segel, 1975).
KS,app J.lmax,app
~ __ ~-L ____________________ p
-~--~------------p
FIGURE 5.11. Estimation of true parameters Ks (a) and I1max (b) from apparent values in
replots of data taken from previous plots in case of competitive (a) and noncompetitive
(b) inhibition.
FIGURE 5.12. Dixon plots for the esti- 1/r

mation of inhibition parameter K 1, re-
spectively rmax and K m , for different
cases of enzyme inhibitions: (a) com-
petitive, (b) noncompetitive, and (c)
un competitive inhibition. a
__~-L~~~~__________ ip
-K
I 1/r
_.... .... ---

5
-- b
5.2.2.2 Enzyme Regulation and Control

In addition to enzymatic reactions, microbial growth phenomena involve the
pronounced ability ofliving cells to adapt to environmental changes. A micro-
organism adapts its activities to changes in the environment by the following
types of mechanisms (Fig. 4.9):
1. Simple regulations by direct mass-action law. These changes are generally
covered by ordinary kinetics (unstructured models). Rapid response is
characteristic in this case, and the time constants of these changes are small.
2. Regulation of enzyme activity by small molecules. These effectors cause
changes in conformation and activity that can be positive or negative. These
so-called "allosteric controls" are vital to the integration of metabolism in
organisms (Monod, Changeux, and Jacob, 1963). The key enzymes in
metabolic networks are especially affected by allosteric phenomena, and
change in kinetic behavior is rapid. The time constant is normally small.
3. Regulation of enzyme amount. Enzyme synthesis is governed by such
regulatory mechanisms as induction, repression, and catabolite repression.
These mechanisms, which may have drastic effects on cellular composition,
have time constants on the order of minutes to hours, and hence result
in a much slower adaption than the mechanisms treated before. Drastic
changes in the kinetic behavior of the cell are the result. These changes may
severely lag behind the changes in environmental conditions. Such "lag
times" in transient modes of operations constitute a "biological inertia" that
must be taken into account in mathematical modeling (Roels, 1978).
If an enzyme molecule binds more than one molecule of S, so that the overall
binding can be represented as
(5.28)
the response of the enzyme activity to S may be sigmoidal rather than
hyperbolic. This may also be interpreted on the basis of an allosteric change
in conformation of the enzyme (Monod, Changeux, and Jacob, 1963).
A mathematical function describing allosteric inhibition (Hill, 1910)
snH
r=rmax KH (5.29)
+ snH
is known as the Hill equation, where nH is the number of substrate-binding
sites per molecule of enzyme, normally in the range of 1 to 3.2 (Hill coefficient),
and KH is a constant comprising the intrinsic dissociation constant Km and
some interaction factors of substrate binding. The definition is analogous to
that for Km (see Fig. 5.13). The estimation ofthe model parameters nH and KH
can be carried out with the logarithmic form of the Hill equation
(5.30)
A plot oflogr/(rmax - r) versus log s (cf. Fig. 5.14) is a straight line with a slope
ofnH and an intercept of -logKH' When logr/(rmax - r) = 0, r/(rmax - v) = 1
and the corresponding position on the log s axis gives log so,s:
SO,5=~ (5.31)
HYPERBOLIC
,,-
SATURATION TYPE ...... _ - - - - - -
...... ,,"
""""""""",
rmax
-2- ---~----------
/1
I
/ "
:
1
1
I I
I I
~~~+----------+---------------- __ c
FIGURE 5.13. Basic biokinetic models including simple hyperbolic saturation type
(e.g., Monod with Ks) and sigmoidal type (e.g., allosteric Hill with KH)'
FIGURE 5.14. Hill plot of all- Ig __r_

osteric kinetics according to rmax - r
Equ.5.30.
~----~--~~-------------------Igs
The influence of experimental conditions on Hill parameter estimation is

described by Glende and Reich (1972). According to Reich et al. (1972), a
nonlinear regression with a minimum of 15 measuring points should be used
for an unbiased determination of parameters. Also, the Hill model has been
extended by Adair (1925) to allow for the existence of stable, partly liganded
intermediates.
It is interesting to note that this sigmoidal character of enzyme kinetics
plays an essential role in regulation. Most of the regulatory enzymes in the
metabolic pathways of amino acid biosyntheses and in carbohydrate metabo-
lism exhibit such behavior, with the result that the kinetics can change in the
presence of effectors (intermediary metabolities) from sigmoid to hyperbolic
(see glycolysis oscillations, Sect. 5.2.1).
The nature of some simplifications is outlined in detail in the articles of
Frederickson et al. (1970), Roels (1978), and Roels and Kossen (1978). How-
ever, the construction of such a model requires the structuring of the biomass
(see Fig. 4.8).
A number of publications have appeared on the dynamics of enzyme
synthesis in a variety of situations. Most of the models are based on more or
less sophisticated versions of the operon model of Jacob and Monod. The role
of m-RNA and its stability were modeled by Terui (1972). Repressor and
inducer control was treated by Knorre (1968), Imanaka et al. (1972; 1973), van
Dedem and Moo-Young (1973), and Suga et al. (1975). Allowance for dual
control and catabolite repression was made by Toda (1976). [See also the
kinetic treatment by Yagil and Yagil (1971), Imanaka and Aiba (1977), and
Bajpai and Ghose (1978)]. A simple structured model was developed by Roels
(1978) showing a combination of the features of the models published. More
recently Toda (1981) reviewed the effects of induction and repression of
enzymes in microbial cultures and their modeling.
The basic concept in the modeling of enzyme synthesis in transient situations
involving a change in environment such as an upward shift in substrate level
is that the level of enzyme concentration in the cells must increase. For this
reason, the structural units responsible for enzyme synthesis (m-RNA in the
ribosomes R) must first be produced.
dr
-=krx (5.32)
dt E
where kE is the kinetic constant and r is the concentration of ribosomes; x is

indirectly a measure of e.
The concentration of ribosomes in a state of balanced growth is a function
of the growth rate p (Maaloee and Kjeldgaard, 1965; Tanner, 1970). Thus,
r ~ p. In this way, the changing number of ribosomes represents a measure of
biological inertia (lag time), which will also affect the growth rate (Romanovsky
et aI., 1974). This is expressed in the following differential equation:
dr 1
- = -(r - r) (5.33)
dt tL
where r is the stationary concentration of ribosomes, the value that will be
reached with a time constant tL (lag time, tL ~ l/p). This simple but adequate
approach uses the concept of a characteristic time (see Sect. 4.2, Fig. 4.5, and
Table 4.2).
5.2.3 CONTRIBUTION OF CHEMICAL KINETIC LAWS TO

BIOPROCESS KINETIC MODELING
In addition to the laws of enzyme microkinetics, the kinetic equations from

chemical reactions based on the type 1 situation shown in Fig. 4.12 also
provide a suitable approach. The power law equations with various reaction
orders, n, differ in the cit relationship of their reaction components. In Fig. 5.15,
the time course of substrate concentration is compared for n = 0, 1/2, 1, and 2
and for Michaelis-Menten enzyme kinetics. The substrate disappearance and
the oxygen utilization in biological waste water treatment may be cited as
realistic examples. For the simple case, integration is possible (Levenspiel,
1972). The integrated solutions for various reaction orders are
n=O Co - c = k o t (5.34)
n = 1/2 2(je;; - Jc) = k 1/2 t (5.35)
n= 1 Inco=klt (5.36)
c
1 1
n=2 - - - = k 2 t (5.37)
c Co
A graphical representation of concentration, the square root of the concen-
tration, the logarithm, or the reciprocal concentration versus time leads to
a linear form from which the slope, k" may be determined.
FIGURE 5.15. Concentration versus time 5

plot for the disappearance of substrate
under different kinetic systems: reaction 50
orders n = 0, 1/2, 1, and 2 and Monod
kinetics (M).
IF===--<
II
I
ONLY I ACTIVE i DIFFERENTLY ACTN

or SPECIES of i
REVERSIBLE REAC- EACH WITH nI
TION with n'" 1 or
,
SECONDARY PLOTS
FIGURE 5.16. Flowchart of systematic analysis of formal kinetics using the shape of
experiments as decision criteria together with the property of invariance I and II,
according to Schmid and Sapunov (1982).
In the application of chemical kinetics, a formal kinetic evaluation method

has been proposed (Schmid and Sapunov, 1982). An operation scheme is
illustrated in Fig. 5.16; it uses two properties of cft curves as decision criteria,
called invariance I and invariance II. These properties concern the linear
transformation capability of first- and second-order reactions. Kinetic curves
with various initial concentrations Ci,O can be superimposed over arbitrary
standard curves (ci,o)s by multiplying ordinates by ratios (Ci, o)sfci, 0 in the case
of first-order kinetics, and by multiplying ordinates in the same way plus

dividing the abscissa by the same ratio in the case of second-order reactions.
As recently outlined, the arbitrary reaction-order concept is still a valuable
method for the formal kinetic modeling of bioprocesses in which the experi-
ments are distorted by physical transport phenomena. These simple pseudo-
kinetic types give interesting hints at the mechanism (A. Moser, 1983a,b).
While a zero-order reaction can be the result of a kLl a limitation in the presence
of a first-order reaction or of multi-S kinetics (overlapping S utilization),
pseudo-first-order kinetics are often the consequence of catalyst poisoning in
side reactions of a GIL process or of internal transport limitations such as in
the case of biofilm processes in which the half-order concept is also adequate
(see Sects. 4.5.2 and 5.8).
5.3 Basic Unstructured Kinetic Models for Growth and

Substrate Utilization (Homogeneous Rate Equations)
In general, unstructured models can be considered a good approximation
when the cell composition is time independent (i.e., in balanced growth) or
when the composition of cells is irrelevant at the industrial scale. In applying
such simplified models, one must be cautious in using extrapolations (see
Sect. 2.3).
Fig. 5.17 shows a summary of growth and S utilization phenomena as
a function of substrate concentration under various circumstances (A. Moser,
1978b). The simple Monod diagram f-L = f-L(s) is selected as the form for
graphical presentation. The cases 1 through 7 can be classified as shown (A.
Moser, 1981; Roels and Kossen, 1978).
FIGURE 5.17. Diagram of the mathematical model describing the relationship between
the specific growth rate Jl (or specific rate of substrate utilization, qs) and the limiting
substrate concentration s (A. Moser, 1978b): 1, Monod kinetics; 2, Monod with S
inhibition; 3, Monod with S inhibition at high X; 4, as 3 with P inhibition; 5, as 4 with
endogenous metabolism; 6, as 4 with sequential metabolism oftwo substrates, that is,
biosorption; and 7, transients such as lag phase.
5.3 Basic Unstructured Kinetic Models for Growth and Substrate Utilization 217
5.3.1 J1 = J1(S): SIMPLE MODEL FUNCTIONS OF INHIBITION-FREE

SUBSTRATE LIMITATION (SATURATION-TYPE KINETIcs)
The present-day theory of microbial growth kinetics stems from, and is still
dominated by, Monod's formulation (1942, 1949) of the function J.l = J.l(s),
given in Equ. 2.54. Also, this relation is a homologue of the Michaelis-Menten
equation: Monod derived it empirically, and thus this is a formal kinetic
equation. The consequence is a different interpretation of the parameters J.lmax
and Ks. The microbial growth rate is
s
J.l = J.lmax Ks + s (5.38)
and the S consumption rate is
(5.39)
These equations are related to each other with the aid of the Y concept (Yx1s ;
see Equ. 2.13).
Several formulas have been proposed as alternatives to the Monod equation,
due to the fact that some failures in applying the Monod equation have come
to light:
- Teissier (1936) in an analogy to the organic growth in length
(5.40)
- H. Moser (1958) in an analogy to allosteric Hill kinetics (see Equ. 5.29, and
also Kargi, 1977)
(5.41)
- Contois (1959) and Fujimoto (1963)

s
J.l=J.lmaxK sX + s (5.42)
Powell (1967) compared these functions and elaborated another equation,

taking into account cell wall permeability, substrate diffusion, and cell size by
a factor KD (see Fig. 5.74):
s
(5.43)
J.l = J.lmax(Ks + K D ) + S
It can be shown that the value of J.l according to Monod kinetics approaches
its asymptote too slowly to be a good representation of experimental data
even in simple cases (cf. Fig. 5.18). The experimental model of Teissier (1936)
and of Blackman kinetics (1905)
1.5
1.0
0-----L----5LO----~---I~OO----~---I~50--___
-J 5
FIGURE 5.18. Comparison of three basic biokinetic models for substrate-limited growth
in a conventional plot of fl versus s: Monod equation (cf. Equ. 5.38), Blackman kinetics
(cf. Equ. 5.44), and the "three-constant (parameter) equation" (cf. Equ. 5.47). (From
Dabes et aI., 1973.)
11 1 S
for s < 2K (5.44a)
Ilmax 2 K
and
~=1 for s > 2K (5.44b)

Ilmax
is generally thought to give a better fit because saturation is faster than with
Monod kinetics.
Dabes, Finn, and Wilke (1973) in a more general approach suggested that
(a) only the upper limit of growth rate is fixed by a single enzymatic step
(rds concept proposed by Blackman in 1905), and (b) at low substrate con-
centrations, more than one step in a series of enzymatic reactions influences
growth rate. These authors analyzed the linear sequence of enzymatic reactions
(5.45)
and, using the concept of qss, they derived a general equation suggesting that
reaction rates in a linear sequence cannot exceed the velocity of the forward
reaction rate rt of its slowest step, because this causes a term [the product
IJ(ri - 11)] to go to zero (rds concept).
Kn n (r_,j-l + 11) Pn n
(r-.n + 11)
s=I
n n KK
i
i,
i- 1
i - 1 (r+,i - 11)
nn Kn
+--K"-n,----
n
(r+. n - 11)
(5.46)
This equation simplifies to Monod's equation if one of the forward reaction

rates is much slower than the others. A more complex situation obtains if two
slow reactions are assumed. In the case that the second is slower than the first,
a "three-parameter equation" results
fl- Ks
s = WK + (S.47)
flmax - fl
If the constant K is very small we simply have the Monod equation, but if Ks
is very small the discontinuous function of Blackman kinetics is achieved.
Model discriminations have shown that in a given confidence region of
experimental error, the rival models cannot be distinguished (e.g., Boyle and
Berthouex, 1974). Model functions with a minimum number of parameters
are therefore preferred.
From the standpoint of chemical reaction kinetics, Kono (1968) and Kono
and Asai (1968, 1969a,b) derived equations for growth and production rate
that include a so-called "consumption activity coefficient," rjJ. The equation is
more flexible than the simple Monod relation, and the growth rate is given by
(S.48)
The consumption coefficient rjJ can be numerically estimated from experimental
data on growth, as shown in Fig. S.19. The concept of a critical cell mass
concentration X eri , is used, and it is determined by numerical methods from
the plot of Fig. S.19. The basic concept of the Kono approach is that reaction
order changes at X cri ' following the general growth equation
dx . . . .
-dt -- k'1 . kJ2 c'1 . cJ2 - k 3 c3 (S.49a)
where c 1 , c 2 , and C3 are the concentrations of limiting substrate, co-substrate,
II III aM IV
\
a XOL-_ _L-~_ _~~_ _L -____~~ b
FIGURE 5.19. Representation of the numerical Kono approach in a cft diagram of

growth (a) and a kinetic plot rx versus x (b). In agreement with different well-known
growth phases I-IV, (cf. Fig. 5.23, lag phase, I; transition, II; exponential, III; and
S-limitation decline phase, IV), and incorporating a linear growth phase in case of
transport limitations (V in case B instead of exponential case A), numerical values can
be taken from the plots to quantify the growth behavior with concentrations at certain
times: Xo = inoculum at to, tL = lag-time, Xc = critical concentration at critical time
when exponential growth is finished and reaction order changes from zero to one,
Xm = maximal cell mass concentration. (Adapted from Kono, 1968, and from Kono
and Asai, 1968, 1969.)
and product. Thus

i = 1 andj = 0 if x < Xeri! (5.49b)
i = 0 andj = 1 if x ;:::: Xeri! (5.49c)
By this numerical approach, the appearance of linear growth phases (char-
acteristic for transport limitation) can also be handled, as shown in Fig. 5.19.
Another approach was derived by Mason and Millis (1976) and is given by
s + n
(5.50)
11 = l1asym -+K + rs s
s s
where l1asym is the value of 11 at which the asymptote of the curved function in
a plot of 11 versus s cuts the line s = 0 (Heidel and Schultz, 1978). The constant
r; is the rate of simple diffusion of the substrate into the cell, given in
[m 3 . kg-I. h- 1 ]. Equation 5.50 applies when transport of substrate across the
cell membrane is permease mediated and when this is the rate-determining
step. Thus, this model describes deviations from Monod growth when the
substrate is at a high concentration, and hence the teim r; .
s is significant.
Schultz et al. (1974) showed that carrier-mediated diffusion through membranes
also obeys a kinetic equation such as Equ. 5.50 with a variant interpretation
of parameters (l1asym = (Dc c)/dm and r; = Ds/d m , with Ds and Dc the dif-
fusivities of substrate and carrier complex in the membrane of thickness dm,
and c the carrier concentration).
Alternatives to Monod-type kinetics for growth and S-utilization models
postulate more structured mechanisms. Verhoff et al. (1972) divided cell
growth into two steps, assimilation and ingestion (including cell division):
Xass + v2 S ~
k2
Xinges! ~ (1 + vdxass + V 3 (V 2 - v1 )S (5.51)
Assuming that assimilation is the rds, the following equation has been
developed for 11 = l1(s):
J-l (k 1 AS k2 + k3)
k3 . VI = - 2k 3 VI + 2k 3 VI
(5.52)
where
ki = rate constants
Vi = stochiometric coefficients
As = concentration difference of limiting
substrate outside and inside cell
Equation 5.52 reduces to simple Monod kinetics under certain conditions and
also to the Lotka-Volterra relationship (cf. Sect. 5.6).
Another approach, quite different from that previously proposed, also

operates with a dual S-utilization pattern, including uptake (assimilation or
elimination from the liquid phase) and growth (degradation or utilization of
S for growth):
(5.53)
where Sa is the structural substrate not available for biomass synthesis and
Se is the functional substrate used for growth (Nyholm, 1976). The growth
rate is expressed here as a function of intracellular concentration of the
limiting substrate (Sint/X) and of "conserved" substrates such as inorganic ions
(phosphate, nitrogen) or vitamins that are not broken down after uptake into
the cells but rather are stored. Thus
(5.54a)
and
d(Sint)
---;{t = 'S,elim - 'S,degrad (5.54b)
Obviously, this model is of practical interest in case of biological waste water

treatment (see Sect. 5.9), because it accounts for the transition between limiting
and nonlimiting conditions. In these situations, Monod-type kinetics are not
applicable to conserved substrates, because fl is controlled by the amount of
intracellular substrate, which may deviate significantly from extracellular
concentration.
Another S-utilization mechanism has been proposed consisting of a con-
sequence of two phenomena (adsorption and penetration of substrate molecules
on specific sites followed by transportation through the cell membrane). In
this approach, the following generalized kinetic model equation was derived
(Borzani and Hiss, 1979; Hiss and Borzani, 1983):
(5.55)
where 'I' and r/J are complex parameters (described in the original paper),
depending on several fermentation parameters. Both parameters can be
calculated from experimental data, even though at present no reasonable
interpretion of individual variations can be given.
The objectives in developing a generalized model are (a) to aid in evaluating
experimental growth data and (b) to aid in determining the conceptual basis
and in suggesting other useful forms to describe microbial growth. From the
general behavior of growth as a function of substrate concentration, a "driving
force" for change in fl with respect to S may be thought of in terms of(flmax - fl).
Thus
(5.56)
where k is the kinetic constant and p is the reaction order (Konak, 1974).
This equation is analogous to the well-known power law concept in chemical
kinetics. It is not implicitly derived from a mechanism, but rather is a purely
formal procedure. Using the concept of "relative" growth rate rrei in a deeper
biological sense (Tempest, 1976)
fl
rrel =-- (5.57)
flmax
Equ. 5.56 can be rewritten
d(fl/flmax)
ds
= k fl~~~ . (1 _~)P flmax
(5.58)
Konak demonstrated that this equation reduces to the simple Monod relation-
ship for p = 2 and to the Teissier form for p = 1. At the same time a quite
interesting relationship between flmax and Ks becomes apparent
1
Ks=--- (5.59)
flmax k
Similar generalizations of the differential form of the specific growth rate
equation have been derived by Kargi and Shuler (1979), combining the differ-
ential forms of the four aforementioned specific growth rate equations. The
following expression was the result of this work:
d(fl/flmax) =
ds
K(~)m.
flmax
(1 _~)P flmax
(5.60)
where K, m, and p are constants. This equation is of the same mathematical

form as the "logistic equation" (Kendall, 1949). Comparing Equs. 5.60 and
5.57 shows, with K = k flmaX' the full identity of both approaches. Table 5.1
gives the values of the constants of the generalized rate equation. The gener-
alized rate equation for microbial growth, therefore, includes most widely used
models as special cases with different constants. Equation 5.60 can be written
as
Inv = InK + mlnv + pln(l - v) (5.61)
where v = drrel/ds; this is a more convenient form for parameter evaluation.
TABLE 5.1. Values of Constants in Equ. 5.60.

Model K m p
Monod IlKs 0 2
Teissier IlKs 0
H. Moser nIK/n I - lin I + lin
Contois IlKs x 0 2
Konak k J1.ma, P P
It is interesting to note that Vavilin (1982) developed a generalized model

function independent of reactor type used for aerobic treatment of waste
water (aeration tank, trickling filter, rotating disk) with a minimum number
of coefficients:
qS=qS.maxKn-p. p+ n (5.62)
S 50 5
Another generalization of microbial kinetics was recently introduced (Levens-

piel, 1980; Han and Levenspiel, 1987).
Finally, the influence of saline environment on the growth of mono and
mixed culture is to be reported and summarized (Esener et al., 1982a). In Fig.
5.20, typical results are shown for the dependence of J1 resp. qo on NaCl
concentration. Mixed culture data where taken from Imai et al. (1979a,b). A
similar decrease was found also in the case of influence of salinity on yield
factors, thermodynamic efficiency and the observed ratio of COD to BOD in
activated sludge processes (Esener et al. 1981). A consequent decrease in Yolx
will than call for a higher aeration capacity. Thus the engineer will be faced
with an optimization problem. In Fig. 5.21 SIX shows the amount of sludge
ri,MAX
x rO
i, r1AX
FIGURE 5.20.Effect of salinity quantified with salt-concentration cNaCI on relative rates

of maximum growth I1max resp. respiration qO.max in pure microbial cultures. (Reprinted
with permission from Esener et a1., Proc. Second IWTU, Vo1. 2, copyright 1982,
Pergamon Books Ltd.)
_1_
YX/S
FIGURE 5.21. Effect of salinity (C NaCl )

on yield coefficients YX[o and YX[S'
showing the optimization problem in
real bioprocessing where both yields
are important at the same time (Esener
10 20
et a1., 1982).
formed per mole substrate consumed and YS[x/YO[X is the amount of oxygen
taken up per mole substrate consumed. Therefore it is desirable to minimize
both. Efficient waste water operation e.g. must be carried out to minimize
sludge production while maximizing yield on oxygen.
5.3.2 f-l = f-l(X): INFLUENCE OF BIOMASS CONCENTRATION ON

SPECIFIC GROWTH RATE
Normally it is assumed that microbial growth follows first-order kinetics with

respect to biomass concentration (cf. Equ. 2.7). In exponential growth in batch
processes and in continuous operation, /1 ;:::; /1max can be realized. However,
Monod (1949) himself presented different hypotheses concerning the function
/1 = /1(s, x), illustrated in Fig. 5.22.
While Verhulst (1845) assumed a linear decrease according to
(5.63)
represented by curve 3, a more realistic function is given in curve 2, quantified
by
So - (x/Y)
/1(s, x) = Ilmax Ks + So - (x/Y) (5.64)
as a result of substrate limitation following a Monod-type kinetics (Meyrath,

1973). Curve 4 represents the improbable case where /1 is inversely proportional
to x
(5.65)
Obviously, the growth rate constant will formally become dependent on

biomass concentration in case of substrate limitation at high cell concentra-
tions. This fact is of great significance in the area of population dynamics.
Using Equ. 5.63 for the calculation of cell mass as a function of time, a kinetic
equation can be derived
110NOD
FIGURE 5.22. Graphical representation

of the formal dependence of specific
growth rate J1 on biomass concentra-
tion x with several approaches from
- X the literature taken as hypotheses.
N. . ,,0 . e.u~.. 1
N - 0 f"'max (5.66)
- Jl~ax + mx ' No(e.u~,,1 - 1)
This is known as Verhulst-Pearl's equation, and it can be used in place of
the normal non-substrate-limited growth of a population
(5.67)
5.3.3 J1 = J1(t): EXTENSIONS OF MONOD-TYPE KINETICS TO

STATIONARY AND LAG PHASE
While the simple Monod-type function is valid only in the exponential and
decelerating growth phases, extensions can be introduced to extend the range
of validity to the lag, stationary, and death phases.
Figure 5.23 gives the complete time course of a discontinuous growth
processes; the growth curve as a function of time is reflected in Fig. 5.1.
Whenever a bioprocess shows an exceptional lag phase, the simple growth
kinetics Jl(s) should be augmented with a time-dependent element Jl = Jl(s, t).
5.3.3.1 Kinetic Modeling of Lag Phases

The duration of a lag phase before the specific growth rate reaches its maximum
value is conveniently defined by the method of Hinshelwood (1946) and Lodge
and Hinshelwood (1943):
tL = t- (2.3)
- - log-
Jlmax
x
Xo
(5.68)
InS
InX x
t So 3
4
5
-t
FIGURE 5.23. Concentration/time plot representing the growth phase of a discontinuous

culture of microorganisms: (1) lag phase with tL = lag time and 11 = 0, (2) exponential
phase with 11 = IlmaX' (3) decay phase with 11 = !(s), (4) stationary phase with 11 = - kd'
and (5) lysis phase (endogenous metabolism phase) with kd > O.
A plot of logx versus t may be successful, and it follows from Equ. 2.6
(Pirt, 1975)
x = Xo . el'(t-td (5.69)
It is known that the lag time tL (cf. App. II) may depend on the size of inoculum
xo, quantified by Hinshelwood (1946) as
const
tL = - - - - (5.70)
No - const
and may be correlated with substrate concentration (Edwards, 1969)
tL = 1.16(8" 0.4) (5.71)
Other formal kinetic equations for the quantification oflag phases in microbial
growth are found in the literature. A simple extension of Monod-type kinetics
using the lag time tL as model parameter is given by Bergter and Knorre (1972):
I1(S, t) = I1max-k
s (1 - e- t / tL ) (5.72)
s+s
The plot of Fig. 5.23 may be used for the determination of I1max and tL. Other
formal kinetic approaches are described by Kawashima (1973) and Edwards
(1969). From a physiological point of view, the lag gives information about
the mechanism of RNA and enzyme synthesis (see also Fig. 4.9). Indeed, the
study of lag times was the basis of Jacob's and Monod's discovery of the
regulation of enzyme synthesis (see Sect. 5.2.2.1). A lag phase is found not only
after inoculation but also in cases of sequential utilization of substrates; this
is called "diauxie." This phenomenon can be quantified by the formal macro-
approach (A. Moser, 1983a) as shown in Appendix II. A mechanistic model
oflag phases has been introduced by Pamment et ai. (1978) and discussed by
Barford et ai. (1982). Starting from the observation that tL was not inoculum-
size dependent, Pamment and colleagues found the major factor involved to
be the availability of the enzymes for glycolysis and respiration. This can be
described by a structured model (Pamment et aI., 1978) with two different
kinds of biomass, but the number of parameters is high, and Equ. 5.72 is more
practical in bioengineering calculations (A. Moser, 1984a).
5.3.3.2 Kinetic Modeling of Stationary Phase of Growth

The logarithmic or exponential law, given by the equation
rx = I1max x (5.73)
can be modified to incorporate a stationary growth phase. This revised form
is known as the "logistic law" (Kendall, 1949; see also Equ. 5.58)
(5.74)
where IY. and f3 are empirical constants.

With ex = J1.max and f3 = X max , La Motta (1976) used the logistic equation for
the quantification of continuous growth cultures. Even though this approach
has been successfully applied in fitting growth curves (e.g., Constantinides
et a!., 1970a and b), the drawback is that there is no clear relationship between
J1. and s in Equ. 5.74. However, the need for this equation is apparent in
situations where product formation in fermentations occurs when microbial
growth has ceased, for example, in antibiotic fermentation and also in bio-
logical waste treatment processes with T-+ high.
5.3.4 NEGATIVE BIOKINETIC RATES-THE CASE OF MICROBIAL

DEATH AND ENDOGENOUS METABOLISM
5.3.4.1 Kinetic Modeling of Microbial Death Phase

The microbial death phase must also be considered, especially in bioprocesses
with long residence times, such as in biological waste treatment where one
might expect an increasing chance of death with an increasing time of exposure.
Various hypotheses exist (Fig. 5.24), including transitions between a strictly
exponential form (curve I) and an abrupt death (curve IV). The Monod equa-
tion, therefore, must be enlarged by including a specific death rate kd[h- 1 ]
in the balance equation for x, defined as follows:
kd = _~.dx (5.75)
x dt
Thus, a first model approach to a reaction S -+ X is
rx = (J1. - kd)x (5.76)
together with
1
-rs = - rx (5.77)
YxlS
This model was successfully used for activated sludge processing (Chiu et a!.,
1972).
In App. II a computer simulation of this model equation in batch processes
(A. Moser and Steiner, 1975) with varying values for kd is presented. The
influence of a neglected kd value on the evaluation of Monod parameters in
linearization diagrams is shown in Fig. 5.25, whereby false or even negative
values for Ks can be the consequence (Grady et al., 1972). With the known
value of kd' it is apparent that the undisturbed values for J1.max and Ks can
be estimated (Moser and Steiner, 1975) only from a plot corresponding to
Equ.5.78
1 Ks 1
-- =_ . _ +1- (5.78)
J1. + kd J1.max s J1.max
However, physiological considerations lead to the conclusion that two

independent factors are responsible for the decline of the overall mass of the
..L
f.l
I
5.24 5.25
FIGURE 5.24. Various hypotheses for microbial death rate: strictly exponential form,
curve I; sudden death after a certain time of exposure, curve IV; or curves II and III
due to inequalities among cells in a population (Hinshelwood, The Chern. Kinetics of
the Bacteria Cell, 1946, Oxford University Press.)
FIGURE 5.25. Representation of the influence of endogenous metabolism (as per Equ. 5.76
with kd ) on obtaining model parameters for microbial growth with Monod kinetics
using a double reciprocal plot. (From Moser and Steiner, 1975.)
biomass. Viable cells are loosing mass due to endogenous metabolism, and
they are also being converted to dead cells. A structured modeling of these
phenomena, therefore, is necessary. Sinclair and Topiwala (1970) have done
so by assuming individual rates of death of viable cells k~ and of endogenous
metabolism k e Thus, as illustrated in the case of a CSTR in Fig. 5.26 in
dependence of dilution rate D
kd = k~ + ke (5.79)
The effect of the true rate of death k~ on the steady-state yields becomes
significant only at very low dilution rates, as shown in Fig. 5.26. Also, an
apparent lag can be explained in batch cultures if the inoculum is of very low
viability.
5.3.4.2 Kinetic Modeling of Endogenous Metabolism

Living cells that are thermodynamically characterized by a state far from
equilibrium must take in high-energy substances and transform chemical
energy to heat. This energy is necessary, for instance, for maintaining osmotic
pressure and for repair processes on DNA, RNA, and other macromolecules.
Thus, a source of energy must be consumed not only for macroscopic reac-
tions like growth but also for maintenance of cellular structure. Therefore, the
substrate balance equation must include a term rns for maintenance (Pirt, 1965)
1
qs = + -
YX/S
f + ms
x qs
A
0.08
" YX/S
0.06
k'
d
5.26 5.27
FIGURE 5.26. Plots of steady-state kinetics in a CSTR by assuming individual rates of

death of viable cells and endogenous metabolism, showing the differences in biomass
concentration x versus dilution rate D in this structured approach. (From Sinclair and
Topiwala, 1970.)
FIGURE 5.27. Diagram of the estimation method for yield coefficient, Yx1s , and ms or k~
of the kinetic model for endogenous metabolism, Equ. 5.80. (Adapted from Pirt, 1975.)
1
-rs = --j1"X + msx (5.80)
Yx1s
together with
rx = fl x (5.81)
where
ms = _(~. dS) (5.82)

x dt e
The balance for energy-source utilization given in Equ. 5.80 can be used for
the estimation ofthe maintenance coefficient m s , plotting rs versus rx , as shown
in Fig. 5.27. It is evident from this graph that
ke
ms=-- (5.83)
YxlS
which means that the specific maintenance rate ke may be regarded as a turn-
over rate of biomass and is useful for comparing the turnover rates of biomass
components with maintenance energy requirements (Herbert, 1958; Marr
et aI., 1963). An alternative plot for a CSTR is represented in Fig. 5.28. It is
interesting to compare the different approaches given with Equs. 5.76 and 5.77
(model 1) and Equs. 5.80 and 5.81 (model 2) to show the model approaches
1/Y FIGURE 5.2S. Alternative graphical method for

the estimation of maintenance coefficient ms
by using chemostat experiments at different
dilution rates D, as represented in the graph.
(Adapted from Pirt, 1975.)
+--------------------1/D
VALUE OF
110DEL
PARAI1ETER I I
______________ IL ___________________________ I I~ODEL 1
+-1_ _ _ _~
kd
I~ODEL 2 _---1..I ___________________________ 1I ___________
1-_:-:-_ ke_
ms I I
I I
REGION 2 I I REGION 1
FIGURE 5.29. Comparison of different model approaches to endogenous metabolism

and maintenance by showing the validity of the different model parameters k d , k e ,
and ms.
used to quantify endogenous metabolism. In Fig. 5.29 the concentration time

curves of a batch process are illustrated.
Even when the parameter kd used in model 1 is, strictly speaking, valid only
in the region where kd can be estimated from the decline in the X curve (region
1), this model has been successfully applied to real bioprocessing (e.g., Chiu
et al., 1972). The reason for this is that the reaction scheme behind model 1
(5.84)
where Xv is the viable cell mass and Xd is the death biomass, indirectly
quantifies at the same time endogenous metabolism (see Equ. 5.83).
FIGURE 5.30. Kinetics of cell death due to the 5

U -jJ (1---)
absence of substrate, and the method for ob- 'd - , d,max Kd 5
taining the parameters f1.d.max and Kd of the
model Equ. 5.85. (Reprinted with permission
from Chern. React. Eng. Adv. Chern. Ser.,
Vol. 109, p. 603, Humphrey. Copyright 1972
American Chemical Society.)
On the other hand, model 2 operating with the parameter ms, which can
be estimated directly from the decline of the S curve in the beginning of the
batch run with rx = 0 (region 2), represents a scheme of parallel reactions:
~x
S (5.85)
~ endogenous metabolism
where ms is assumed to be constant over the whole range of the cit curve.
Other modifications of the basic models for endogenous metabolism have
been proposed by Humphrey (1978) and by Gutke (1980, 1982). These show
an S-concentration dependence of the parameters fld and ms as given in
Equs. 5.86 and 5.87 and Fig. 5.30:
(5.86)
ms -m
-
s )
S,max ( Ke + S
(5.87)
The most probable model will include both terms kd and ms, thus combining
Equs. 5.76 and 5.80.
A comparison of unstructured growth models of microorganisms is given
by Takamatsu et al. (1981), and Esener et al. (1983) discuss the applicability.
5.3.5 KINETIC MODEL EQUATIONS FOR INHIBITION BY

SUBSTRATES AND PRODUCTS
In most cases of inhibition, the formal kinetic model equations are, like
Monod's relationship, derived from theories of the inhibition of single enzymes.
The equations are, however, only hypotheses; they may be replaced by any
other adequate model.
5.3.5.1 Substrate Inhibition Kinetics

There are a large number of formal kinetic equations in the literature:
1. Andrews (1968,1969,1971) and Noack (1968) analyzed substrate inhibition
in chemostat culture using
1 8 1
Jl= Jlmax l + K/ (5.88)
s 8 + 8/K ',S ~Jlmax'-K
S + 8
'1 + 8/K ',S
where K"s is the substrate inhibition constant.

2. Allosteric substrate inhibition (Webb, 1963)
8(1 + 8/Ks)
(5.89)
Jl = Jlmax 8 + Ks + 82 /Ks
3. Multiple inactive enzyme-substrate complexes (Yano et aI., 1966)
1
(5.90)
Jl = Jlmax 1 + Ks/8 + L (8/K"s)i
j
4. Empirical function (Aiba et aI., 1968)

8
/I = /I ---' e -s/KI,S (5.91)
r' r'max Ks + 8
5. A semiempirical form of the Teissier equation
Jl = Jlmax(exp -.:
',S
- exp - ~)
S
(5.92)
6. An equation derived by Webb (1963) assuming the Debye-Hiickel theory

for the effect of ionic strength, 0', on the rate
8
/I_II e1.17a (5.93)
r - rmax 8 + Ks(1 + O'/K"s)
Edwards (1970) compared Equs. 5.89 through 5.93 with experimental results
from the literature and found that discriminating among them was not possible.
He therefore recommended the first form, Equ. 5.88, as the simplest of these
equivalent equations. The behavior of Equ. 5.88 is shown in Fig. 5.31 and
as a computer simulation in App. II.
The estimation of parameters is indicated in Figs. 5.31 and 5.32. K"s is the
value of 8 when half the maximum rate Jl is reached. While Fig. 5.32 is adequate
for an initial estimate, the three equations in Fig. 5.31 provide more accurate
estimates (Humphrey, 1978). Other methods include a rearrangement of
Equ. 5.88 for 8 Ks leading to
1 1 1
-=--+ '8 (5.94)
Jl Jlmax Jlmax' K"s
which can be used to determine K"s'
,..."" ""
.... '" 'Monod
1/2 ...../'"/
"
"".::"
/' "'} /PIPmax
flmax " '" ..
2VYI'
1...L..~--------1/S
) l' 1/KI
5.31 5.32
FIGURE 5.31. Kinetics of substrate inhibition and method of parameter estimation by

precise method using computer simulation of the kinetic equation (cf. Equ. 5.88).
(Reprinted with permission from Chern. React. Eng. Adv. Chern. Ser., Vol. 109, p. 603,
Humphrey. Copyright 1972 American Chemical Society.)
FIGURE 5.32. Kinetics of substrate inhibition and method of parameter estimation by
approximation using a double reciprocal plot based on Equ. 5.88.
A quite different equation for the quantification of substrate inhibition

kinetics has been proposed by Tseng and Waymann (1975) (Waymann and
Tseng, 1976). Substrate concentrations higher than a characteristic threshold
substrate concentration Sc inhibited growth linearly in accordance with
s
Ji = Jimax Ks +s - KI,s(s - sc) (5.95)
where K),s is the inhibition constant. This approach should be used in cases
where the inhibition pattern does not have the shape of the curve shown
in Fig. 5.31 but rather shows a linear decrease, as schematically shown in
Fig. 5.33 (Kosaric et aI., 1984).
A generalized concept of a rate equation for one-substrate enzymatic reac-
tions was proposed by Siimer (1978) for situations where the substrate Sand
products Pi can inhibit the reaction. The following general basic reaction
mechanism was suggested as a basis for the derivation of a rate equation:
SE ~ SES ~SE+PI+P2
1~ pKs 1~
EP I ~ E ~ {ES} ~E+ PI + P2 (5.96)
H 1~ ""K P2 1~ K P2
EP I EP2 ~ EP2 ~ EP2S ~ EP2 + PI + P 2
According to this scheme, SES and EP2 S complexes are supposed to be

1a 3 2a 2
'--_.1..-_-'--_-'--_..... 5 -p
5.33 5.34
FIGURE 5.33. Comparison of two essential substrate inhibition models according to

Equ. 5.88 and Equ. 5.95 (from Kosaric et aI., 1984) in a diagram of specific growth rate
f1 versus substrate concentration s.
FIGURE 5.34. Possible hypotheses for the effect of product concentration p on microbial
growth rate!1= linear decrease (1), first-order decrease (1a), sudden stop (3), and decrease
after a period of no effect (2,2a). (Hinshelwood, The Chern. Kinetics of the Bacteria
Cell, 1946, Oxford University Press.)
productive. PI inhibits competitively and Pz noncompetitively. Using the

concept of qss with PI = P2 = P, the rate equation as a function of substrate
conversion (s (see Equ. 2.45) with parameters used in the scheme of Equ. 5.96
has the following form:
kcat 'e o(1- (s)[1 + b(1 - (s)/K~ + e(s/KpJ (5.97)
r= a + b(s + c(~
where a, b, and c are functions of So and the dissociation constants K s, K p"
and K p2
5.3.5.2 Product Inhibition Kinetics

Hinshelwood (1946) distinguished different types of concentration-action
curves (Fig. 5.34): linear decrease, exponential decrease, or a stepwise function.
When there is no threshold concentration, the effect is often given by a simple
linear relation (Dagley and Hishelwood, 1938)
S
f-l(s, p) = f-lmax Ks + S (1 - kp) (5.98)
where k is the kinetic constant.

Beyond this numerical fit to product inhibition, which was modified by
Holzberg et al. (1967)
fl = flmax - kl (p - k 2) (5.99)
and by Ghose and Tyagi (1979) in the form
fl = flmax(l -~)
Pmax
(5.100)
for quantifying ethanol inhibiton of yeast growth, the function fl(S, p) can be
modeled by analogy to enzyme kinetics. Jerusalimsky and Neronova (1965)
represented the dependence of fl on P by hyperbolic or sigmoidal curves, and
Jerusalimsky (1967) recommended the following function
S K),p
fl(S, p) = flmax Ks +S KI,P +P (5.101)
where KI,P is the product inhibition constant. This model was used for a
computer simulation of growth in discontinuous culture (Bergter, 1972), and
it represents the most commonly used equation (Aiba et aI., 1968, 1969a,b;
Fukuda et aI., 1978; Peringer et aI., 1974). At the same time, Aiba and Shoda
(1969) developed an empirical approach for the product inhibition pattern
of yeast
S
fl(S, p) = flmax Ks + s e- kp (5.102)
with k the empirical kinetic constant.

A quite flexible class of models has been described by Ramkrishna et aI.
(1966, 1967) using a sequence of reactions in which intermediary products
inactivate viable cells. Oscillations of growth can be quantified using this
model (Knorre, 1976), and a model of this type was used by Fishman and
Biryukov (1974) for growth in penicillin fermentation. A similar approach
is used in general analyses of growth patterns subject to the influence of
intermediates (Knorre, 1980; Petrova et aI., 1977).
Finally, Bazua and Wilke (1977) compared different approaches to ethanol
inhibition of continuous yeast cultures and concluded that observed differences
in the results are caused not only by different strains but also by altering
experimental conditions. They proposed a three-parameter equation
flmax = flo - kl . P(k2 - p) (5.103)
where p is the average value ofp in continuous culture and kl' k2 are empirical
constants. This equation shows that there is a limiting concentration of p,
beyond which the cells grow or produce. These authors also used a two-
parameter of an equation similar to Equ. 5.100 to fit experimental data
- )1/2
(1 + ~
flmax = flo (5.104)
Pmax
with Pmax the concentration of P at which cells are still viable. This equation
shows some similarity to the generalized Monod equation developed by
Levenspiel (1980), with a Monod-type equation replacing flmax by
kobs k ( 1 - -P- )n (5.105)

Perit
where
k= flmax
Perit = critical product concentration at which fermentation ends

n = toxic power number
To unravel the interacting effects of the four rate constants of the complete
rate equation with Equ. 5.105 (k, Ks, Perit> and n), it is more convenient to use
a CSTR than a CPFR or a batch reactor. Making a series of runs at different
P values, a plot can be made as shown in Fig. 5.10b, resulting in the parameter
values of Ks and k (1 - P/Perit). Then, with the value of Perit easily estimated
in advance, k and n can be found from a plot, as given in Fig. 5.35. The specific
nature of the full rate equation with Equ. 5.105 says that there is a definite
upper P limit for inhibitory effects (Perit) above which fermentation ceases.
The empirical constant n accounts for the fact that inhibitory effects may
work in different ways: linearly (n = 1), rapid initial drop followed by a slowing
rate to Perit (n > 1), or vice versa (n < 1), as shown schematically in Fig. 5.36
(Levenspiel, 1980). Experiments often show these types of inhibition effects
(e.g., Hoppe and Hansford, 1982). Recently, Han and Levenspiel (1987) pro-
posed a basic approach to biokinetic modelling by extending the Monod-
equation to inhibition by substrate, product and biomass.
find J-I from this intercept!

IgJ.l max ---------~~----~
II
I
I
too I
I
low I
I
guessed:
'Pcrit :
too high I A P
o Ig (1- p/Pcrit) Perit
5.35 5.36
FIGURE 5.35. Evaluation of the toxic power coefficient n in the generalized microbial
kinetic equation according to Levenspiel (1980). See Equ. 5.105.
FIGURE 5.36. Decrease ofthe observed rate constant in the generalized Monod equation
modified with Equ. 5.105 with increase in the inhibiting product concentration p,
showing the effect of toxic power number n. (From Levenspiel, 1980.)
5.3.6 KINETIC MODEL EQUATIONS FOR REPRESSION

Catabolite repression plays a central role in many industrial fermentations such
as in yeast technology (diauxic growth) and secondary metabolite produc-
tions. Even though the exact nature of the biochemical mechanisms involved
in these regulations is often unknown (see Demain et aI., 1979), the following
type of formal kinetic equation
1 s 1
r- = r (5.106)
IX I,max Ks + s(l + s/K R ) '-x
can be used as adequate approach. This equation is a formal analogy to the
case of S inhibition (cf. Equ. 5.88 and Fig. 5.30) and was used for modeling
yeast diauxie (A. Moser, 1978c; 1983a) and antibiotic production (Bajpaj and
Reuss, 1981; Schneider and Moser, 1986; Moser and Schneider, 1988). Kinetic
equations of structured models for induction and repression have been reviewed
by Toda(1981).
5.3.7 fl = fl{pH)
The influence of pH of a bioprocess can be dealt with in combination with
inhibition kinetics. Andreyeva and Biryukov (1973) gave a number of different
models for dealing with combined inhibition (see also Brown and Halsted,
1975).
For example, using Equ. 5.88 a formal kinetic equation can be constructed
wherein the H+ ion concentration h+ is treated as if it were a substrate
concentration. The value of the constant K is found by a method analogous
to that in Fig. 5.29 using the concentration at the half-maximum value of Jl.
from the two parts of the curve representing stimulation K 1 and inhibition
K2 (Humphrey, 1977b, 1978). Thus, a mixed inhibition function is obtained
s
r.(s pH) = r (5.107)
I , "max(Ks + s)(l + Kdh+ + h+ /K 2 )
Andreyeva and Biryukov also gave a numerical curve-fitting procedure
similar to Equ. 2.58 for evaluating the pH dependence of a rate
rj = rj,max(OC O OC 1 pH OC 2 ' pH2 ... ) (5.108)
with OC j the coefficients of a polynomial.
5.3.8 KINETIC PSEUDOHOMOGENEOUS MODELING OF MYCELIAL

FILAMENTOUS GROWTH INCLUDING PHOTOSYNTHESIS
Active growth of molds occurs only at the tips of the hyphae. Bergter (1978)
presented a kinetic model for calculating the relative rates of apical growth
and branching of mycelial microorganisms in submerged cultures by measuring
J.l and the quotient LIN with L the total length of hyphae per unit of culture
volume and N the number of hyphal tips per unit of volume. On the basis of
the simplest case of Hinshelwood's network theory (Dean and Hinshelwood,
1966), the following equation may be used to quantify the growth of small
mycelial trees of Streptomyces hygroscopicus on solid surfaces
and (5.109)
where k1 is the mean relative rate of apical growth (J.lm h -1), and k2 is the
mean relative rate of branching (J.lm -1 . h -1).
It can be concluded that
(5.110)
The different dependence of k1 and k2 on s is shown in Fig. 5.37. With
increasing s, the density ofthe colonies, which depends on the branching rate
k2' decreases faster than the radial growth rate; this can be interpreted in
connection with the regulation oftransport of metabolites within the hyphae.
Cellular differentiation and the connection to product formation in molds was
elaborated from Megee et al. (1970).
The growth of fungi has been reviewed by Prosser (1982), and growth of
photosynthetic microorganisms has been summarized by Aiba (1982) and by
Pirt et al. (1983) together with production in tubular photobioreactors. An
unstructured model of algal growth has been developed by ten Hoopen et al.
(1980) based on the Monod equation.
5.3.9 KINETIC MODELING OF BIOSORPTION

The phenomenon of "biosorption" occurs in biological waste water treatment
("sludge adsorption," Jones, 1971; Walters, 1966), in which substrates are
eliminated rapidly from the liquid phase due to "adsorption" and "storage"
in the sludge, but degradation itself lags behind the initial storage. Other
examples of biosorption can be found in the literature pertaining to con-
ventional fermentations (Bayer and Meister, 1982; A. Moser, 1974). The
phenomenon can be detected by measuring the heat of combustion of sludge
(differential thermal analysis). Simultaneous utilization of both light energy
and substrate is another type ofbiosorption (Follmann, Markl, and Vortmeyer,
1977).
Typical experimental findings are shown in Fig. 5.38 (Theophilou, Wolfbauer,
and Moser 1978). Elimination is quantified by COD measurements, while
degradation is given by the BOD values, occurring with a time delay. The
difference is the effect of biosorption.
Quantification and modeling of biosorption must account for the biphasic
nature just mentioned. Pseudo homogeneous treatments of biosorption fail
because pseudohomogeneous macrokinetics are represented by a zero-order
equation
0.02
0.01
_s L----"''''----_t
5.37 5.38
FIGURE 5.37. Dependence of kinetic parameters kl and k2 in the model of filamentous

growth (see Equ. 5.110) on glucose concentration s in case of chemostat culture of
Streptomyces hydroscopicus. (From Bergter, 1978.)
FIGURE 5.38. Kinetics of adsorption (biosorption) in biological waste water purifica-

tion. Time course of chemical and biological oxygen demand expressed as eliminated
substrate Se and degraded substrate Sd following the reaction scheme of substrate
degradation and substrate elimination, and evaluation ofthe rate of adsorption Sa from
the difference in chemical and biological rates (Theophilou et aI., 1978). The final level
SI represents the undegradable substrate. The value of Sma. represents the maximal
capacity of cells (sludge) to "adsorb" substrate (phenomenon of "biosorption").
FIGURE 5.39. Consequences of

o-ORDER
",
....Zq
....... -- MONOD TYPE
KINETICS
"biosorption" effect for kinetics of .... 5

biological waste water treatment ads-'elim
/' COD
shown in a plot of specific con- /
sumption rate qs versus substrate / 1SlORDER KINETICS
concentration s with the resulting gmax s

zero-order kinetics. COD
rS,eff = ko,app' x (5.111)

The appearance of this zero-order rate (BOD with respect to S) is seen in
Fig. 5.39 (A. Moser, 1974). Several authors have developed a formal kinetic
approach to biosorption effects (Busby and Andrews, 1975; A. Moser, 1977;
Theophilou et aI., 1978). Figure 5.40 illustrates the reaction scheme thought
to be adequate as a starting point for the following set of model equations:
(5. 112a)
(5. 112b)
r ads = r elim - r degrad (5.112c)
GAS LIQUID SOLID PHASE
l
G:---'------ L I -I
elim ~ degrad I
i ~L _:.cx7-'-'---1"OO-\:(~S +~ Os) X .. X + CO2
I .
I :
I I REACTION !!
S~lim : Sads-Sdegrad MEASUREI1ENTS

I ~COD NBOD I . BOD "PSEUDOHOMOGENEOUS APPROACH"
FIGURE 5.40. Elucidation of the "biosorption" phenomenon in a reaction scheme

including substrate elimination (elim) quantified by COD removal, and degradation
(degrad) accompanied by O 2 utilization (BOD) and "adsorption" (ads) following this
pseudohomogeneous approach to the LI S process.
Adsorption rate often follows the approach given in Equ. 5.50. Thereby
Selim,O = Sdegrad,O - S1 - Sads (5.113)
and
k(x' Sdegrad,O' ~ads)
= Sads (5.114)
A _ _ A
S - Smax
1 + k(x' Sdegrad. 0 ~ads)
Equation 5.114 is formulated by analogy to the adsorption isotherm (Langmuir
kinetics, see Equ. 2.55) using the sorption capacity ~ads as measurement for
sludge activity
8
~ads = 1- -A- (5.115)
smas
(8 = adsorbed concentration, S1 = un degradable S level). The term inside the

parentheses in Equ. 5.114 can be regarded as the sorption potential. The value
of smax is typical for a given waste water and is about 200 kg- 3 at x = 5 kg- 3 .
It is interesting to note that Equs. 5.l12a and 5.l12b represent simple first-
order reaction rates; this fact will be explained later as a formal kinetic
approach to multisubstrate reactions (cf. Sect. 5.5). The most essential feature
of such biphasic biokinetics is that they are able to explain the observed
experimental discrepancies by incorporating nonidentical kinetics in the liquid
and solid phase.
5.4 Kinetic Models for Microbial Product Formation

5.4.1 METABOLITES AND END PRODUCTS
According to Gaden (1955, 1959), four types of product accumulation can
be distinguished on the formal level based on the quantitative relationship
5.4 Kinetic Models for Microbial Product Formation 241
CD
FIGURE 5.41. Concentration/time plot of the three basic types of microbial product
formation: (I) growth association. (II) mixed growth association, and (III) nongrowth
association (Gaden, 1955).
between the amount of product and the growth of cells. These types can be
observed in Fig. 5.41, which is a cft plot.
Type O. Type-O production as a supplementary case occurs even in resting cells
that use only a little substrate for their own metabolism. The microbial cells
function only as enzyme carriers. Steroid transformation and vitamin E
synthesis by Saccharomyces cerevisiae are examples.
Type 1. Type-1 situations include processes in which product accumulation is
directly associated with growth; this is the case for primary metabolites, in
which the formation of the product is linked to the energy metabolism.
Examples include fermentation to produce alcohol and gluconic acid (Koga
et aI.. 1967), and situations in biological waste water treatment.
Type 2. Type-2 processes include fermentations in which there is no direct
connection between growth and product formation and also no direct or
indirect link to primary metabolism (secondary metabolites), for example,
penicillin and streptomycin.
Type 3. Type-3 processes include those having a partial association with
growth and thus an indirect link to energy metabolism. Examples include
citric acid and amino acid production.
Recognition of the type of production is done with the aid of plots of rdt
or qdt, as shown in Fig. 5.42 for types I through III. The most significant plot
is given with specific rate qj.
The diagram of the time dependence of the specific rate of a bioprocess is
called the "quantification diagram": It gives the best insight into the process
and is basic for designing mathematical models (cf. Sect. 2.4.2.1).
Product formation linked to microbial growth can be described by
rp = YP1x rx (5.116a)
qp = YP1x1l (5.1 16b)
where YPlx is the product yield referred to biomass formed. If product yield is
expressed in terms of substrate used we have (e.g., Constantinides, 1970a,b
(5.117)
FIGURE 5.42. Formal kinetic diagrams of the three basic types of microbial product
formation (I-III) expressed as volumetric rates 'x, 's, rp (a) or as specific rates 11, qs,
qp (b), according to Luedeking and Piret (1959).
II
!----t,....:::....-III
IV
V
~--------------~~
FIGURE 5.43. Formal kinetic linear relationships between specific rates of production
qp and growth 11 exhibiting a different form: (I) growth association, (II) mixed growth
association, (III) nongrowth association, (IV) negative correlation, and (V) no corre-
lation. For equations see text.
Hence the following relationship exists between yield factors

Y
y Pls = Y Plx (5.118)
XIS
Substituting a Monod-type equation into Equ. 5.117 results in a hyperbolic

function for production in the case of growth association
s
r - q (5.119)
P - P,max Ks + S 'X
Non-growth-linked product formation (curve III in Fig. 5.43) is more
difficult to quantify because no direct relationship to growth exists. As an
alternative in this case, the dependence ofrp on biomass concentration is often

successfully used
(5.120)
Product formation of this type can also be quantified by the dependence of
substrate utilization (see Equ. 5.117). Similarly, the O 2 concentration 0L is
useful in, for example, the quantification of penicillin synthesis (Giona et al.,
1976)
qp = kl . 0L + k2 (5.121)
When product formation is partly growth linked and partly independent
of growth, a combination of Equs. 5.116 and 5.120 is valid
qp = YP1x'll + kp (5.122)
as proposed by Luedeking and Piret (1959).
In addition to product formation types 0 through III, product formation
types IV (and V) of Fig. 5.43 can also occur when there is a negative (or no)
correlation between qp and Il. This occurs, for example, in melanine production
from Aspergillus niger (Rowley and Pirt, 1972). This type of production can
be modeled by
(5.123)
Terui (1972) also proposed a kinetic model for enzyme formation in the
nongrowing phase
qp = qP,max'exp[ -k2(t - t max )]
(5.124)
where
kl' k2 = rate constant
K 1 = empirical constant
t max = time of maximum production rate qp,max
The general form of Equ. 5.122, with Equs. 5.120 and 5.116 as boundary
cases, suggests a logistic equation (Luedeking and Piret, 1959). This has
been similarly included in the generalized concept proposed by Kono and
Asai (1968a-c, 1969a-c, 1971a,b) using the consumption coefficient rp as
an apparent coefficient of growth activity. The value of rp in each phase of
fermentation is (cf. Equ. 5.48):
Induction phase rp=O (5.125a)
Transient phase rp=rp (5.125b)
Exponential growth phase rp = 1 (5.125c)
Declining growth phase rp = (

Xm
~ Xc ) em X- x) (5.125d)
TABLE 5.2. Types offermentation processes classified according to values of

production rate constants, k P1 and k p2 , in general formula of production rate
(Equ.5.126).
Description Case
+ + Product fonnation associated with growth and nongrowth la and lb

+ o Product fonnation associated with growth 2
o + Product fonnation associated with nongrowth 3
+ Product fonnation associated with growth and decreased with 4
nongrowth
In a declining growth phase, the value of ifJ decreases from unity to zero. When
a constant growth phase is included, the value of ifJ in this phase is expressed
as follows:
Constant growth phase ifJ = Xd (5.125e)

X
Xd is the cell concentration at the boundary of an exponential growth phase

and a constant growth phase. The general equation for product formation
following the Kono concept, thus, is
(5.126)
These authors described the various instances of growth-associated, or growth-
independent, product formation using the parameter k pi ;;e: 0 as given in
Table 5.2. The value of parameters kpi may be obtained in a purely numerical
way directly from the experimental curves, as shown in Fig. 5.44. Case 1b
(see Table 5.2) is included in Fig. 5.44a: It represents a linear growth phase
and shows the similarity with the fermentation pattern described previously.
The relationship between the Kono equation and the formal concept given in
Fig. 5.41 becomes obvious in considering exponential phase with ifJ = 1. For
this situation the following equation results (see Equ,. 5.48):
1
rp = kP1 ' - - ' rx (5.127)
J-lmax
Recently, Asai and Kono (1983) presented a modified equation and applied
it to industrial fermentation processes.
When the product accumulation rate is influenced by the product decom-
position rate, a term rp must be added to general Equ. 5.122 to account for
this event:
(5.128)
Decomposition rate was used, for example, for penicillin production modeling
(Constantinides et aI., 1970a).
In bioprocesses in which rp is directly coupled to energy metabolism, a
generalized treatment also seems to be possible. On the basis of ATP pro-
FIGURE 5.44. Schematic representation of the numerical Kono approach to microbial

product formation expressed as the general formulas of the rates of growth rx and
production r p , including different growth phases (1, induction; 2, transient; 3, exponential;
and 4, declining), according to Equ. 5.126 and Table 5.2. (Adapted from Kono and
Asai, 1968a-c, 1969a-c): (a) both kP1 and kP2 have a positive value. The dotted lines
take into account a linear growth phase, as shown in Fig. 5.19. (b) k P1 > 0, kp2 = O.
(c) kP1 = 0, kp2 > O. (d) kP1 > 0, kp2 < O.
duction per amount of S converted by energy-producing pathways

rATP = lA\Pls' rS,energy (5.129)
and by using a known concept for the relationship between the rate of sub-
strate consumption due to material requirements for biomass precursor syn-
thesis, the following equation results:
rS,synth = Ysrx' rx (5.130)
This result is a type of equation for microbial production already given in
Equ. 5.122 (cf. Roels and Kossen, 1978):
1
rp=------rx
Ysrx +-x
ms
(5.131)
a p ' Yx1s ap ap
with a p the stoichiometric coefficient for product formation. Equation 5.131

states that, even in this case, a partly growth-associated term and a biomass-
associated term both exist.
The difficult modeling of non-growth-associated product formation has
been attempted by introducing time dependence. Shu (1961) proposed an
empirical approach
(5.132)
on the basis of the assumption that rp of individual cells is a genetically

determined function of cell age A. Product concentration at arbitrary time,
therefore, is given by
p = ft x(A) fA z;, kl,i' e- k2 . i A dA dA (5.133)

o 0 I
A mean cell age A(t;) was later defined by Aiba and Hara (1965) as
A(t.) = Xo' Ao + J:~ X' dr (5.134)

I x(t;)
with to ::;; r ::;; t i The product formation rate can be quantified by

(5.135)
where the specific production rate as a function of A can be evaluated directly
from experiments. Fishman and Biryukov (1973) applied this concept for
modeling the production of secondary metabolism. The value of qp (A) can be
determined from the slope of a plot of qp against A, resulting in a model as
given with Equ. 5.108.
Non-growth-associated production formation is the most difficult type to
model. Another simple and yet plausible equation for such modeling involves
the concept of the "maturation time" (Brown and Vass, 1973). The maturation
time is taken as a kind oftime delay between growth and production formation;
this is shown in Fig. 5.45a. The equation is either
(-ddtP) = Yp!x' (dX)

t
-
dt t-t"
(5.136a)
r.
I
a b
~-----'--- X
( g/l )
FIGURE 5.45(a). Graphic representation of the maturation time concept according to

Brown and Vass (1973) in a plot of rates ri versus time, with the possibility of a gross
evaluation of the parameter t M (b) Evaluation of the model parameters tM and Yplx in
the maturation time concept with the help of a graphical "trial-and-error" method
(Brown and Vass, 1973).
or
(5.136b)
In a graph of the product concentration versus the concentration of cell
mass, x, the line is straight when one has made the correct choice of t M ; the
slope of the straight line represents the carrying coefficient. A graphical "trial
and error" method of solution is given in Fig. 5.45b.
A systematic approach to unstructured modeling of microbial production
has been presented by Ryu and Humphrey (1972) using a mechanism similar
to that used for the derivation of Equ. 5.131. Here supplementary branching
at a common intermediate Ii was considered. Assuming that the conversion
of Ii to both cells and product is limited by the rate of one single enzyme
in the chain, and assuming that both processes follow Monod-type kinetics,
the following relationship holds:
qp = qp,max 1 + (8 - 1) / (5.137)
/l /lmax
In this equation 8 is the ratio of the Michaelis-Menten constants of the enzy-
matic reactions leading to growth and to product formation (8 = KJ,;/Ki,i)'
The implications of Equ. 5.137 are graphically presented in Fig. 5.46 (Roels
and Kossen, 1978). These authors indicated that for 8 > 1, a Monod-type
equation similar to Equ. 5.119 applies; for 8 = 1, an equation similar to
Equ. 5.44 applies; and for 8 < 1, one has a parabolic relationship.
Another generalization of product-formation kinetics based on mechanistic
background but still using the formal kinetic approach has been presented by
Bajpaj and Reuss (1980a, 1981). This model, (cf. App. II) along with rate
equations for /l, qs, and qo, was successfully used to simulate experimental
data. The concept of this model approach is to eliminate S between the
equations for /l (see Equ. 5.38) and for qp (see Equ. 5.106). The result is an
equation relating qp with /l in a more structured manner:
[1 - (Jl/Jl...,)] (Jl/Ilmax) (5.138)
qp= qP.....
[(Kp/Ks)'x] [1 - (Jl/Jlm .. )]2 + [1- (Jl/Jlm.. )] (P/Pmax) + (p/Jlm .. )2/[(K.JKs)-x]
Plots of Equ. 5.138 are illustrated in Fig. 5.47 for different sets of values of
the terms [(Kp/Ks )' x] and [(KJ/Ks )' x]. The dotted curves (A) represent
a modification due to inhibitions of growth (Moser and Schneider, 1988;
Schneider and Moser, 1986).
5.4.2 HEAT PRODUCTION IN FERMENTATION PROCESSES

Heat production represents one of the significant macroscopic process variables
(see Fig. 2.3). Experimentally, heat evolution can be determined with the aid
of measurements of temperature changes in the reactor and the volumetric
heat Hv [J/l] estimated according to the Uhlich approximation (see Equ. 3.70).
An interpretation of the kinetic term (dh.,/dt)r in Equ. 3.69 is directly related
OL-__L -_ _ ~ _ _~ _ _ ~
o 0.4 -plJ.Jmax
5.46 5.47
FIGURE 5.46. Graphical representation of the theoretical non-linear relationship be-

tween product formation and microbial growth according to Equ. 5.137 in a qp versus
J-l plot. (Roeis and Kossen, 1978, after Ryu and Humphrey, 1972.)
FIGURE 5.47. Kinetic pattern of secondary metabolite productions in a normalized qp

versus It plot based on repression function (see Equ. 5.106) according to Equ. 5.138,
with variation of the values K[ and Kp (Bajpaj and Reuss, 1982) in curves 1 through 4,
by inclusion of other terms for inhibition of growth due to substrate and product K[sx
and K IPX (Moser and Schneider, 1988) in curves A (~~~):
Kp/Ksx KdKsx
Curve 1 0.0013 0.6667
2 0.013 0.6667
3 0.013 6.667
4 0.0013 6.667
A Same values as in 500 10
curves 1-4 200 10
100 10
to fermentation heat analysis (cf. Sect. 3.3.8). Basically, the production of heat
in a microbiological process can be dealt with quantitatively by the same
equations as product formation. The validity of the following special equation
has, however, been established: It includes a term for growth, one for product
formation, and one for endogenous metabolism (Cooney et aI., 1969; Mou and
Cooney, 1976) using the specific rate qH.
1 dhv 1 1
qH, = _.dt" = -y:-. p, + y;-. qp + mH , (5.139)
x XIH, PIH.
where YXIH , and YP1H, are the thermal yields related to biomass and product
formation [gjl] and mH , is the maintenance coefficient [J jg . h]. The evaluation
of Y and qH, is done in a way parallel to that shown in Fig. 5.27.
A simplified equation is possible for the rate of heat formation for aerobic
processes. This equation utilizes the rate of oxygen utilization as a basis

(Cooney et ai., 1969):
1
qH, = y-qo (5.140)
OiH,
This procedure for estimating heat evolution is simpler than the previous
method. The proportionality constant (l/YOiH ) for this correlation is inde-
pendent of ji, slightly dependent on the substrate, and possibly dependent on
the type of organism growth. This property of the constant (l/YoiH ), which is
thought to be identical with the reaction enthalpy of fermentation, ~HkO)
[MJ/mol 02J related to oxygen consumption,
1
--=~HkO) (5.141)
YOiH,
makes it extremely valuable when growth is to be carried out on complex
substrates. According to Minkevich and Eroshin (1973), ~HkO) varies from
0.385 to 0.494 MJ /mole for filamentous fungi and from 0.385 to 0.565 MJ/mole
for bacteria (Luong and Volesky, 1980), following the descending order bacteria,
yeast, mold.
As mentioned previously (Sect. 3.3) in connection with the rate of heat
transport, the reaction enthalpy ~HkX) or ~H1f) is given in units of [kcal/gX]
or [kcal/gSJ for practical reasons. Table 5.3 summarizes ~HR for fermenta-
tions (Bronn, 1971). A systematic treatment is possible using classifications of
aerobic and anaerobic processing versus substrate; the numerical values are
independent of the strain of microorganism.
For calculating heat balances, one should note that ~HR values are multiplied
by a process rate and must therefore be in the correct dimensions
dHv
r H, =-=~H'r
dt R x
([kJ/I' hJ = [kJ/g] . [gil' hJ) (5.142)
Fermentations are principally free energy-yielding processes (~G < 0) and/or

are exothermic, ~H < O. At present but few data are available in the literature.
A quite different approach in evaluating the heat evolution in bioprocessing
uses the heat of combustions of the components involved in the conversion
(Imanaka and Aiba, 1976). Molar enthalpies can, in part, be taken from
thermodynamic tables or, in part, calculated by, for example, using the obser-
TABLE 5.3. Metabolic process enthalpies for fermentations.

Process -I1HftX) [kcal(gX] -I1H~') [kcal(gS]
Hexoses~X 2.6 1.4
Hexoses --> P +X 1.5-2.7 0.117-0.162

Hydrocarbons ~ X 6.8 6.8
vation of Kharash (1929) that each mole of oxygen consumed in combustion

results in a release of 0.444 MJ. The following equation can be written in
the case of aerobic batch cultures of Sacch. cerevisiae for the rate of heat
produced:
rH , = (-AHs)~ + (-AHN)JlX _ (-AHx)JlX _ (-AHp)qp.x

Ms Yxls Mx Mx Mp
(5.143)
where r H , is the rate of heat offermentation [J. h- 1 .1- 1 ] and -AHs, -AHN'
- AH x' - AHp are the heats of combustion of substrate, ammonia, cells, and
product [J mol- 1 ]. Rearranging Equ. 5.143 results in an expression for the
heat of fermentation per unit mass of cell produced AHkX ) [J. g-l of cell]
(Voleskyet aI., 1982)
(5.144)
where
-AHN + AHx
K1 =------ (5.l45a)
Mx
-AHs
K z =--- (5.1 45b)
Ms
-AHp
K3 =--- (5.145c)
Mp
Similarly, heat of fermentation per unit mass of substrate consumed AH~)
[J. g-l of substrate] can be expressed as
-AH~) = (K1 + Kz -
K3 qp) Yx1s (5.146)
Yxls Jl
Similar results were obtained by recent work with microcalorimetry (Brettel
et aI., 1981a,b; Lovrien et aI., 1980; Oura, 1973).
Imanaka and Aiba (1976), using the concept of heats of combustion, came
to the same result as that expressed in Equ. 5.140, by direct measurements of
the evolution of volumetric heat Hv [J .1- 1 ].
5.5 Multisubstrate Kinetics

In real situations, there are complex cases such as in biological waste water
treatment and fermentation technology (complex media with multiple carbon
sources, e.g., molasses, worts, metabolic intermediates, vitamins, etc.) that
cannot be treated with the simple model equations Jl = Jl(s) given in Sect. 5.3.
In the course of a growth process in a complex medium, the valuable, easily
utilized components are exhausted after a short time. For use of the remaining
5.5 Multisubstrate Kinetics 251
FIGURE 5.48. Multisubstrate kinetics for a 5

two-substrate reaction: (a) Concentration/
time diagram for strictly sequential substrate
utilization (diauxic growth). (b) Partly over- 1+2
I
lapping and partly sequential substrate use. 1" ....
(c) Simultaneous substrate utilization. Sum-
type kinetics are often applied in (b) and (c).
a
components, the cellular enzymes responsible for their breakdown must first
be synthesized. The cells pass through many transitions; a series of growth
phases will result, each with a successively decreasing growth rate.
Figure 5.48 summarizes practical situations in bioprocessing (A. Moser,
1981). Classification of situations can be achieved by distinction between
sequential and simultaneous utilization, with a transition case of overlapping
utilization. While sequential or consecutive consumption of substrates (Monod,
1942, 1949) can often be analyzed in two separate growth phases, the simul-
taneous utilization encountered in biological waste water treatment is more
difficult for mathematical modeling.
5.5.1 SEQUENTIAL SUBSTRATE-UTILIZATION KINETICS

A general equation for sequential substrate utilization appears to be given by
the relation
(5.147)
where the factor irepr is a formal expression for catabolic repression operative
on the use of S2 as long as Sl is present in the medium (see Equ. 5.106). A
similar equation for the diauxic growth according to Equ. 5.147 is given by
Imanaka et a!. (1972) resp. A. Moser (1978b) where the simple relationship
(5.l48a)
resp.
1
frepr(sd = 1 + s/K R (5.148b)
is used to represent regulation (cf. Equ. 5.106). Bergter and Knorre (1972) used
another function for this purpose
irepr(S1,t) = L(K~: s - frepr) (5.149)
that incorporates a diauxic lag time t L2 This is a formal analogy to the

formulation of biological inertia of Romanovsky et al. (1974), in which growth
rate is related to ribosomal concentration CR , where cR is a stationary value
dC R 1_
dt = tL (c R - cR ) (5.150)
The opposing inhibitory effects of the substrate in catabolism are taken

into consideration by an interaction term in the kinetic equations (Aris and
Humphrey, 1977; Knorre, 1977; Yoon, Klinzing, and Blanch, 1977)
Jl. = Jl.1(S1,S2) + Jl.2(S2,S1) (5.151)
This case will be outlined in Sect. 5.5.3.
Strictly sequential S utilization is encountered, for example, in beer brewing,
where glucose, maltose, and maltotriose are used one after the other (Budd,
1977). A simple model describing the kinetics of wort sugar uptake during
batch fermentation of brewers' wort by yeast can be found in the literature
(Fidgett and Smith, 1975); it operates with the concept of a critical concentra-
tion used as a switch from one limiting substrate to another, and it uses
a simple "on-off" concept for individual rates. Basically, sequential utilization
operates with only one limiting substrate at a time, expressed as an equation
with additive terms (see Equ. 5.147). Sometimes, however, terms appear
multiplicative in the final equation. This discrepancy will be discussed in
Sect. 5.5.3.
Another concept for a formal kinetic description oftwo-substrate utilization
with partial overlapping uses the idea of a critical substrate concentration
(A. Moser, 1978b, 1981) that is directly measurable (W6hrer et a!., 1982).
The value sl,crit is the concentration at which there is, for a short time,
simultaneous utilization of S1 and S2 so that the factor
1
/1=---- (5.152)
1 + St!Sl,crit
can be substituted in Equ. 5.147 instead of frepr.
5.5.2 SIMULTANEOUS SUBSTRATE-UTILIZATION KINETICS

The case of simultaneous uptake is common in biological waster water treat-
ment and elsewhere, and the following equation represents a general approach
(Wuhrmann, von Beust, and Ghose, 1958)
(5.153)
or, using enzyme kinetics (Atkinson et aI., 1969; Wilderer, 1976),
qS,lol = L:i (qS,max'i K S,is~ Sj ) X (5.154)
in which Ks,i is a constant for the ith reaction and qS,max,i is the contribution
of the ith reaction to the maximal rate such that (Shehata and Marr, 1971)
qS,max = L:i qS,max,i (5.155)
The appearance of an overall reaction order one as the sum of several uptake
rates was explained by such approaches (Wuhrmann et al., 1958), and it was
later quantified (Wolfbauer et aI., 1978). This fact is indicated in Fig. 5.48c.
However, it seems to be possible to use the simple Monod equation for
modeling, and an increased value of apparent Ks will be the consequence.
A case of parameter estimation with double substrate limitation is shown in
Fig. 5.49 using the Walker plot (see Fig. 4.21b) for the integrated form of
the rate equation (Wilderer, 1976). Difficulties arise due to overlapping and
to the phenomenon of biosorption (cf. Sect. 5.3.9), leading to a zero-order
behavior even if the overall order of S utilization is unity (see Equ. 5.36).
Other groups have also worked successfully with similar first-order summary
kinetics to represent the components in biological waste water treatment, Si
(for example, biological oxygen demand, BOD, or chemical oxygen demand,
COD) (Eckenfelder and Ford, 1970; Grau, Dohanyos, and Chudoba, 1975;
10schek et aI., 1975; Krotzsch et al., 1976; Oleszkiewicz, 1977; Tucek et aI.
1971). In the relationship due to 10schek et aI. (1975)
-rs = ksSXf(T) (5.156)
FIGURE 5.49. Plot of two-substrate

kinetics in a Walker plot (cf. Fig. 1 Send-SO
' - - - - - - - - - - - - In -==--
4.21b). (Wilderer, 1976) t-to Send- S
the rate constants show summary kinetics with the following relationship to
the kinetic parameters (Moser and Lafferty, 1977)
ks = Jlmax , [l/g. h] (5.157)

Yx1s Ks
where the endogenous metabolism is indirectly reflected in an elevated K~
value (Moser and Steiner, 1975). The equation due to Grau et al. (1975)
-rs=k.X(:Jn (5.158)
with the integrated form for the case ns =1
S = So exp ( - k S: .t) (5.159)
uses a correction with variables a and b for metabolizable and nonmetaboliz-

able substrate in the form
S = BOD _ b (5.160)
a
and is in agreement with the empirical relationship of Tucek, Chudoba, and
Madera (1971) for the case ns = 2.
To a first approximation, Equ. 5.159 can be written
S kx t
-=1--- (5.161)
So So
Eckenfelder and Ford (1970) used the equation
SO=I+kXt (5.162)
S
whereas Krotzsch et al. (1976) expressed the relationship between substrate
and cell mass with
(5.163)
and used a Contois kinetic approach.
5.5.3 GENERALIZATIONS IN MULTISUBSTRATE KINETICS

Generally, in multicomponent media, substrates will be utilized sequentially.
Physiologically, two different responses can be distinguished, known as cata-
bolite repression and inhibition, both resulting in a diauxic growth pattern.
Remarkable differences in growth behavior can be observed comparing batch
and continuous operation in a CSTR (Harder and Dijkhuizen, 1975), due to
the existence of much lower steady-state concentrations of Sj in CSTRs so that
the effects of both repression and inhibition may be absent or reduced.
Yoon et al. (1977), Aris and Humphrey (1977), and Knorre (1976) derived
generalized Monod equations for multisubstrate systems, based on the fol-
lowing sequence of reactions:
(5.1 64a)
(5.1 64b)
X/~2X (5.164c)
X"~2X (5.164d)
where X' and X" are different intermediary states of the cells and ai is
the stoichiometric coefficients. Applying the steady-state approximation, the
following equation results
with a complex meaning of parameters KSl and KS2 according to the

mechanistic interpretation of Km (cf. App. III).
Further I1max.l = k3 and I1max.2 = k 4 The stoichiometric coefficients are
given by
(5.1 66a)
and
(5.1 66b)
and indicate that each substrate exhibits a competitive inhibition effect toward
the utilization of other substrates. Noncompetitive inhibition was considered
by Lee (1973). For n multiple substrates, Yoon et al. (1977) derived Equ. 5.167
from these considerations
(5.167)
and applied such kinetics to batch and continuous two-substrate situations.

Another generalization of the Monod-type kinetics was suggested by Tsao
and Hanson (1975) and by Tsao and Yang (1976). They assumed the existence
of growth-enhancing substrates SI and of essential substrates SE, resulting in
an equation
(5.168)
The effect of growth-enhancing substrates is given as the sum of Monod-type

expressions, while essential substrates form a product. The value of Jlmax will
be the sum of Jlmax, 0 and all Jlmax,i' When all SI are missing, a value of Jlmax,o
is still possible. By this approach a type of equation can be explained by
assuming glucose and Oz as essential substrates.
With an aerobic bioprocess it is evident that, in addition to substrate, Oz
can be rate limiting. The curve for biomass concentration increasing with time
in a batch run is significantly influenced by the oxygen transfer rate quantified
by the volumetric transfer coefficient kLl a. This situation of an external
transport limitation is depicted in Fig. 5.50 and in App. II (cf. also Sect. 4.5.1).
Kinetic modeling in this case is successful with the aid of a double-substrate-
limitation function (Megee et aI., 1972; Ryder and Sinclair, 1972) as shown by
Reuss and Wagner (1973)
S 0
Jl(s,o) = Jlmax Ks +S Ko +0 (5.169)
This, equation was incorporated into the mass balance equation for Oz where
oxygen transfer was also considered
1
ro = kL a(of - od - - - ' Jl(S, 0)' x (5.170)
Yx10
Equation 5.169 can be used as a formal kinetic approach without assuming
a mechanism.
The question that generally arises is whether more than one substrate can
exert control in a given system. Bader et ai. (1975) and Bader (1978,1982)
showed that for the growth rate to be able to respond to a controlling substrate
or to both substrates simultaneously, the stoichiometric line must intersect
the transition line 0( = /3 (sdKs! = sz/Ksz ). The stoichiometric line relates
the steady-state values of the substrates, which line, rearranged in terms of
dimensionless substrate concentrations, become in dimensionless form
R _ /3- _
PO -
Yxls\' Ks! (0(0 -
-)
0( (5.171)
YXIS2 ' Ksz
This is shown in Fig. 5.51. For this intersection to occur, the following
inequality must be satisfied
Ks! /30 1 (5.172)
YXls\--'Ks2 >->
YX1S2 0(0
This is in agreement with the requirements proposed by Sykes (1973) and

defines the shaded region in Fig. 5.51, indicating that both double-S limita-
tions and a switch between limiting substrates are quite common. However,
Equ. 5.172 holds only for kinetic models such as the Monod and exponential
models approaching saturation asymptotically. To handle this problem, it is
desirable to overlay curves of constant Jl on Fig. 5.51 for some of the models
discussed earlier. Two different philosophies can be distinguished: interactive
and noninteractive models.
An interactive model is based on the assumption that if two substrates are
x
A
E.....l------:------slKs1
5.50 5.51
FIGURE 5.50. Linear growth as a consequence of transport limitation in case of O2

quantified with kLl a (after Reuss and Wagner, 1973) in a concentration/time graph.
FIGURE 5.51. Dimensionless plot showing the feed conditions required for double-
substrate limitations to be possible (shaded area). The stoichiometric line must cross
the transition line with sdKs1 = s2/Ks2 (Reprinted with permission from In Microbial
Population Dynamics, Bader, 1982. Copyright CRC Press, Inc., Boca Raton, FL.).
S2/KS2
lOA
~::max
9
8
7
6
5 -
0.75
4 Monod Kinetics
0.5 C
00 I 2 3 4 5 6 7 8 9
FIGURE 5.52. Conceptual representations of the interactive model. (a) Sl is converted

to P1 by an enzyme that requires S2 as a cofactor. (b) Substrates Sl and S2 from two
parallel pathways are combined by enzyme Er to produce a product P1 that is required
for growth. (c) Plots of lines of constant dimensionless specific growth rate P,/P,max as
a function of two dimensionless substrate concentrations for interactive models of
the Megee type (cf. Equ. 5.169) with Monod kinetics (Reprinted with permission from
In Microbial Population Dynamics, Bader, 1982. Copyright CRC Press, Inc., Boca
Raton, FL.)
present in less than limiting concentrations, both substrates will affect the
overall rate. The simplest model is constructed by simply multiplying two
single-S-limited models together: Figure 5.52a shows the conceptual repre-
sentation of such an approach.
For strictly sequential substrate utilization, Bader (1978) has proposed

a "noninteractive" model that will seldom be encountered in its purest form.
A noninteractive model basically implies that 11 is limited by only one substrate
at a time. Therefore, the growth rate will be equal to the lowest rate that would
be predicted from the separate single-S models. For Monod-type kinetics, this
would be written as follows:
81 81 8z
11 = Ilmax, 1 K + 81 for --<-- (5.172)
SI KSI Ksz
82 82 81
11 = Ilmax,Z K + 82 for --<-- (5.173)
S2 KS2 KSI
Figure 5.53a represents the concept of this approach, examples of which may
be found in the literature (Ryder and Sinclair, 1972; Sykes, 1973).
Comparisons of both approaches for Monod kinetics are shown in Figs. 5.52c
and 5.53b. Noninteractive models are, by their nature, discontinuous func-
tions at the transition line from one substrate limitation to another, predicting
L~/flmax
10
b 9
8
0.85
Monad KinetiCS
'----------0.75
L - - - - - - -_ _ _ 0.5
0 I 2 3 4 5 6 7 8 9 10 51' KS1
FIGURE 5.53. (a) Conceptual representation of the noninteractive model. Systems 1
and 2 operate independently of one another. (b) Plots oflines of constant dimensionless
specific growth rate 11/l1max as a function of two dimensionless substrate concentrations
for noninteractive models of the Megee type with Monod kinetics (Reprinted with
permission from In Microbial Population Dynamics, Bader, 1982. Copyright CRC
Press, Inc., Boca Raton, FL.)
5.6 Mixed Population Kinetics 259
higher values of J1 in the region where sl/K s1 and S2/KS2 are small. Interactive
models are continuous functions but may err on the side of predicting lower
values of J1 when tI. and f3 both are small. It seems unlikely that any two cellular
subsystems would be totally independent of each other, even though the
degree of interaction may be rather small. Both types exist for certain types
of substrates. Finally, it has been shown by Bader (1982) that the region of
simultaneous limitation by two substrates is extremely small, so that double-S
limitation can be regarded as a rare event, difficult to realize. Nevertheless,
a closed analytical solution of multi-S kinetics is needed for engineering
calculations.
5.6 Mixed Population Kinetics

Mixed populations of microorganisms (Aris and Humphrey, 1977; Jannasch
and Mateles, 1974; Yoon and Blanch, 1977) occur, for example, in waste water
treatment, in which the variation of organism composition with time plays
a central role. The connection between various environmental circumstances
(such as substrate choice and organism species composition) is shown in
Figs. 5.54 and 5.55.
Even in natural self-purifying systems like rivers, a succession of popula-
tions occur: chemoorganotrophic bacteria, ciliates, nitrifiers, algae, and fish
(cf. Fig. 5.54). In biological waste water treatment reactor systems, a plurality
of consecutive reactions is known as a result of activities of the biocoenosis,
shown in Fig. 5.55 (Wilderer and Hartmann, 1978). Mixed popUlations are
found in aerobic and anaerobic sludge digestion; beer, corn, cheese, and yogurt
production; the human skin; the alimentary canal; aquatic environments; and
the soil. A similar but simpler situation is found in the case of nitrification
or denitrification of biological waste water: Both have been described as
a two-step microbial process (Eggers and Terlouw, 1979; Tanaka et aI., 1981).
Another case of mixed population interactions is the rumen fermentation
(Czerkawski, 1973), for which mathematical models have also been developed.
The initial supply of organic substances in waste water plants is utilized
especially for carbon-containing components by bacteria (population X d as
shown in Fig. 5.55. The nitrogen-containing components in a conventional
operation are utilized later by nitrifying (population X 2) and denitrifying
(population X 3 ) cells. The phosphorus-containing compounds are stored in
particular populations due to multiple interactions between aerobic, anoxic,
and anaerobic zones. Bacteria also are themselves food for bacteria-consuming
organisms such as protozoa and ciliates. These are all coupled in a food chain,
population Xi' collectively referred to as the "biocoenosis" or ecosystem.
In analogy to natural systems, some technical processes involving mixed
popUlations have recently been introduced. In a situation where the optimal
strategy for the production of ethanol from sugars involves batch fermentation
of highly concentrated sugar solutions, the use of dual organisms that possess
FIGURE 5.54. Relative predom-

inance of microbial organisms
utilizing organic substances in
waste water: sarcoidia (XI),
holophytic flagellates (x z), holo-
zoic flagellates (x 3 ), bacteria
(x 4 ), ciliates (xs), suctoria (X6)'
stalked ciliates (X7), rotifers (Xg).
_ .. 5
,, FIGURE 5.55. Reaction with a
,
...
mixed population of micro-
\
,, organisms in a low-loaded waste
water purification operation as
\
a function of the time course of
environmental changes (mixed
substrate, SC.N.P)' X I, sapro-
phytic bacteria; Xz, nitrification
bacteria; X 3 , denitrifying bac-
teria; X 4 to Xi' organisms in the
food chain. (Adapted from
Wilderer and Hartmann, 1978.)
different substrate and product inhibition characteristics can bring about

improved ethanol productivity. The extent of any improvement depends
crucially on the functional form of the inhibition relationships and on the
values of the kinetic parameters (e.g., Jones and Greenfield, 1981).
5.6.1 CLASSIFICATION OF THE TYPES OF MICROBIAL

INTERACTIONS
Various terms are used to denote various types of interactions; however, there
is so much overlap in the meaning of these terms that they do not fit all
categories of the interaction pattern (Bungay and Bungay, 1968; Meers, 1973;
Noack, 1968). Table 5.4 shows a classification scheme of all combinations
of interactions known, that is, competition, commensalism, amensalism,
TABLE 5.4. Classification of pairwise interactions between microbial populations

based on the signs of the entries aji and a ij from the community matrix A (i = j)
(May, 1973).
Effect of species j 011 species i (sign of a'j)
o +
--: competition - 0: amensalism - +: predation
Effect of species i
on species j o 0-: amensalism 00: neutralism 0+: commensalism
(sign of aj ,)
+ +-: predation + 0: commensalism ++: mutualism
mutualism, and predation. The influence of species ion speciesj is determined

by defining an element a ij in the so-called "community matrix" (Bailey and
Ollis, 1977). If aij is positive, species j has a positive effect on growth of species
i; an inhibitory effect is described by a negative value of a ij .
Competition occurs when a community of two or more species are mutually
limiting because of their joint dependence on a common factor external to
them. Commensalism is the case where the growth of one species is promoted
by the presence of a second species in a population, the growth of the second
species being unaffected by the presence of the first. Mutualism is similar to
commensalism, but both organisms grow faster in the presence of the other
than they do separately. It could be caused by the production of growth factors
or products that serve as nutrients. Symbiosis is similar and occurs if the
mutualistic partnership is necessary for survival of one species. Synergism is
a third type of mutualism in which the formation of specific products is greater
in mixed than in pure cultures. Amensalism is the situation where the growth
of one species is repressed because of the presence of a toxic substance pro-
duced by another; it represents the opposite of commensalism. Neutralism,
which is relatively rare, means that the two species have no observable effect
on one another. Predation occurs when an organism totally engulfs and
digests another organism, which thereby loses the ability to reproduce itself.
Parasitism, which for microbial interrelationship is difficult to distinguish
from predation, occurs when one organism feeds or reproduces at the expense
of tissues or body fluids of another.
Supplementary to this classification scheme, a distinction is often made
between open and closed environments, that is, continuous and discontinuous
reactors. An open environment leads to population stability (homeostasis),
although oscillations may be observed due either to periodic fluctuations in
the environment or to interactions between microbial species. The factors
influencing the survival of given species are different in closed and open
systems (Meers, 1973).
5.6.2 KINETIC ANALYSIS OF MICROBIAL INTERACTIONS

The behavior and mathematical model of a mixed population are basically
similar to those of a pure culture. The Monod model can often be used as
a global approach, even though the kinetic parameters Jlmax, K s , and Y

are not strictly constant. Examples are biological waste water treatment
(Gaudy and Gaudy, 1972) and biogas fermentation (Andrews, 1971; Chen
and Hasimoto, 1978). However, several distinct steps can sometimes be dis-
tinguished, for example, in the biogas process the fermentative, acetogenic
and methanogenic reaction. Thus the following scheme is the starting point
for the setup of a mathematical model:
Si ~ fatty acids + H2 + CO2 ~ acetic acid + H2 + CO2 ~ CH 4 + CO2
(5.174)
5.6.2.1 Competition (Between Two Organisms Growing on a Single

Substrate)
With the simplest model including two species (X l ,X 2 ) competing for the same
limiting substrate (s), the principle of selection can be demonstrated ("survival
of the fittest" -organism with the greater specific growth rate). The scheme
can be written
S.....-AXl (5.175)
--"X2
The kinetic model, using the Monod equation for Jli = f(s), will be for a CSTR
r X, = Jll(S)Xl - DXl (5.176)
r X2 = Jl2(S)X 2 - DX2 (5.177)
1 1
rs = ---Jll (s)x l - --Jl2(S)X 2 + D(so - s) (5.178)
Yx,1s Yx21s
Figure 5.56 shows the steady-state values of this model as a function of D.

Evidently no coexistence is possible. However, for lower dilution rates (D < Ds)
the first species survives, whereas for a middle range of D (Ds < D < Dc)
the second species survives. Above the critical dilution rate Dc both organisms
wash out. The plot of the kinetics D = Jll es) and D = Jl2eS) indicates that the
switch from the first surviving species to the second is due to the intersection
of both Jl characteristics (Gutke, 1980, 1982). This model is the basis for studies
of ecosystems and of the problems of the evaluation or selection of mutants
in a CSTR for genetic optimization studies.
The principle factors governing the kinetic behavior are the differences in
Jl and Ks of both species, for example, mutants at different dilution rates.
The dependencies Jl(s) can be divided into two main categories. The first type
(cases I, II, IV, V in Fig. 5.57) is characterized by the fact that the Jl(s) curves
for both mutants have no common point apart from the origin. In the second
type (cases III, VI), the two curves intersect (see Fig. 5.57). It is clear that the
curves will cross each other if Jlmax,2 < Jlmax, land KS2 < K Sl ' All of these cases
are illustrated in Fig. 5.58, where correspondingly denoted areas reflect the
magnitude of the ratios Jlmax, d Jlmax.2 and Ksd KS2 (Sikyta et aI., 1979). Thus,
0.4
's
o
______________________~>o
0.5 0.5
l ~ "
"
"
d "I:
",
I
" ,
__ b
-' '
_~_"''''''2S
_---- -]- I X,~O
0~~-='-~---=-~~--~--~~_~~1~1~'==~0
o 0.5
5
bc --D
FIGURE 5.56. Steady-state concentrations of two competing microbial species x! and
Xl and of the limiting substrate s in dependence of dilution rate D in a CSTR according
to Equs. 5.176 through 5.178. Parameter values for simulation: Jimax.! = 0.8 h -I,
KSI = 0.01 g '1- 1 ; Jimax.l = 1 h- I ; KSl = 0.05 g ,1- 1 ; YX21s = 0.5. For explanation see
text.
FIGURE 5.57. Theoretical dependencies of specific growth rate Ji on the limiting substrate
concentration s for a population including two microorganisms (e.g., mutants). A first
type (cases I, II, IV, V) exhibits no common point apart from the origin, while a second
type (cases III, VI) shows an intersection between both curves. (From Sikyta et al.,
1977.)
Pmax,1 FIGURE 5.58. Generalized

flmax,2 plot of functions from Fig.
flmax,1 KS1
5.57 with indications of
~max,2 = KS2
ranges of cases I through VI.
III
(Sikyta et aI., 1979).
I-------i"---- JJmax,1 =fl max,2
IV
I.-_ _ _ _ _L -_ _ _ _ _ KS1
KS2
with known values /lmax and K s, Fig, 5,58 can serve for the identification of
a given trend and for the prediction of population changes. The dilution rate
can be used as a controlling factor, The point of intersection in Fig, 5.58 can
be found by putting /ll equal to /l2 and solving for s:
/lmax, 2 KSl - /lmax,l KS2
Sintersect = (5.179)
/lmax, 1 - /lmax, 2
A chemostat will operate stably at D = /linterseet such that both species will
coexist. With D < /linterseet the situation is the same as before, with species 1
washing out. With D > /linterseet, species 1 will grow faster and species 2 will
wash out.
Growth of two competing species in a closed environment was analyzed by
Volterra (1931), who established many of the fundamentals of mathematical
ecology. Powell (1958) and H. Moser (1957) have provided mathematical
analyses of the fate of contaminants or of mutants in CSTRs. For the survival
of competing cells, the surface/volume ratio is also important, as it affects
the value of /lmax but not of Ks (Veldkamp, 1975). Several papers concern
mixed-culture steady-state studies in the chemostat (e.g., Jannasch and Mateles,
1974; Veldkamp and Jannasch, 1972) and also the dynamics (e.g., Aris and
Humphrey, 1977; Lee et at, 1976). Stability analyses show that possible steady
states depend on the relative disposition of the two growth curves and on
the position of the point whose coordinates are the nutrient feed concentration
and dilution rate. Qualitative phase portraits are drawn for each of the 31
distinct types of situations. The operation of a periodically forced CSTR with
two competing populations was carried out to give criteria for the stability
of the resulting cycles and to show conditions under which stable periodic
trajectories of coexistence can be achieved (Stephanopoulos et at, 1979).
Similarly, a CSTR with cell recycle of heterogeneous populations was ex-
amined with the result that cell recycle can change the outcome of species
competition. Selective recycling of one species can reverse this outcome or
stabilize coexistence by its selective effect on cell residence time (Weissmann

and Benemann, 1979). Other trends to be reported are the multivariable
feedback control of a competing mixed-culture system (Wilder et aI., 1980),
the computer-aided analysis of mixed cultures (Ohtaguchi et aI., 1979), the
mathematical description of competition between two and three species under
dual S limitation in a CSTR (Gottschal and Thingstad, 1982), and the anslysis
of a two-stage CSTR system as an attractive alternative (Stephanopoulos
and Frederickson, 1979). The advantageous use of multistage systems was
recommended earlier by Veldkamp and Jannasch (1972). The complexity of
microbial interactions often observed, however, needs a much more elaborate
kinetic modeling, including categories in addition to competition. By varying
the concentration of the medium components, various interacting systems can
be created (e.g., Tseng and Phillipps, 1981).
5.6.2.2 Commensalism and Amensalism

Many microbes produce substances that promote the growth of other species;
this phenomenon is probably the most widespread form of commensalism.
It follows the scheme
SI -4 Xl p (5.180)
Sn -4 X 2
The combinations of possible commensal relationships, however, are legion,
and the interactions may well be competitive as well as commensalistic, but
are rarely studied in full detail. When during the growth of one species the
environment is altered in such a way that another species is inhibited, due
either to the removal of essential nutrients or to the formation of toxic
substances, published data do not necessarily specify the type of antagonism
or amensalism that has occurred. The toxic products likely to cause amensalism
may be divided into inorganic and organic compounds. Often, the negative
growth rate of species i is caused by the pH change from the growth of
speciesj. Kinetic model equations for pure commensalism without inhibition
employ Monod-type relations for Il; = !(8;) and the assumption that the
nutrient for the dependent species is produced with no loss of yield from
the nutrient for the independent species. The following balance equations for
a CSTR were derived by Reilly (1974):
rx , = (Ill - D)Xl (5.181)
rX2 = (1l2 - D)X2 (5.182)
(5.183)
rs = DS 2 + IllX l _ 1l2 X2 (5.184)

2 YXl YX2
For this simple system, three different steady states are possible.
The problem of a commensalistic model with feedback inhibition and

activation by produced substances was also treated by Reilly (1974). Inhibition
was assumed to be competitive
Sj
Ilj = Ilmax,jKs, j + Sj + K s, j' I/K (5.185)
j I, j
and activation to be additive

S a
".="r-max,j.K
r-J
S,j
+ Sj +"r-max, A K A,j J+ aj
J
j
(5.186)
with a being the concentration of activator and i = j.

This complicated commensalistic system is shown to be less stable, with
limit-cycle response occurring after dilution rate changes, as would be ex-
pected from a model with more parameters. Though virtually any feedback
from the dependent to the independent species will cause some overshoot
after a step change in D, the most pronounced oscillatory behavior is caused
by feedback inhibition and feedforward activation. Limited agreement with
experimental data was obtained even though the analysis was somewhat
limited due to the complexity of the system and the large number of differential
equations. Similar results were obtained by Sheintuch (1980), who examined
the dynamics of commensalistic systems with self- and cross-inhibition. Multi-
plicity of steady states was observed as well as oscillatory states with these
complex kinetics and in the case of a reactor with biomass recirculation.
Stability and dynamics are summarized in a qualitative phase plane by this
author.
Population dynamics and stability analyses in the case of commensalism
and other interactions are widespread in the literature (e.g., Lee et ai., 1976;
Miura et ai., 1980). A modified approach to commensalistic modeling in the
case of yeast growth and nicotinic acid production employed Monod-type
functions for species 1 and 3 but the following equation for 112 (Tseng and
Phillips, 1981):
S2 P (5.187)
112 = Ilmax,2 KS2 + S2 Kp +P
with p being the concentration of nicotinic acid. This paper concluded that
mixed cultures are very complex, and the interacting mechanisms may not be
as simple as commensalism or competition or both. Equation 5.185 was also
used in modeling pure commensalism as single interaction of the type given
in the scheme of Equ. 5.180, with the result that the five eigen-values are all
real and negative, and therefore the steady-state solution is a stable model
(Miura et al., 1980).
5.6.2.3 Mutualism
Mutualism, sometimes called "protocooperation" if it is not obligatory for
survival, is the case if both species grow faster in the presence ofthe other than
they do in pure culture. It can be caused by the production of growth factors

or products Pi that serve as nutrients:
...---- P2 "---- (5.188)

8, ..... Xl.............. ~X2 +- 8"
P1
Following this scheme, the specific growth rates (Ill and 1l2) can be expressed by
Sl P2
III = Ilmax.1K81 + Sl K + P2 (5.189)
P2
and
S2 Pl
112 = Ilmax.2 K 82 + S2 . K Pl + Pl (5.190)
The balance equations in the case of mutualism in a CSTR are

(5.191)
(5.192)
(5.193)
1
r p2 = qp2' x 2 - --Illxl - D P2 (5.194)
Yl /P2
As in the case of commensalism, the steady-state solution is a stable node

(Miura et ai., 1980). When competition is accompanied by mutualism (and/or
commensalism), two kinds of steady-state solutions are obtained owing to the
values of system parameters, one being a stable node and the other a stable
focus or an unstable saddle point. Damped oscillations tend to occur when
either Xl or X 2 has a lower growth ability than the other.
In a somewhat more complex case, Xl grows on 8, which is toxic at high
concentrations to X 2 , and X 2 produces a nutrient for Xl' It may be modeled
as follows by using the same biomass balance as before but different 8 balances
(5.195)
rp = D P + Qpx2
IllX l
- -- (5.196)
YxlP
with kinetic equations for 8 inhibition and co-utilization of P (cf. Equ. 5.190)
(Meyer et ai., 1975). Taguchi et ai. (1978) simulated this model and received
good agreement with experiments. The model behavior, determined with pa-
rameters taken from the pure cultures, demonstrates the coexistence of both
species at lower dilution rates, above which the singular point of coexistence
could not be obtained (the sole survivor being the strain that overcomes the
competitor). It seems generally valid that a mutualistic interaction in a CSTR
will be established only over a limited range of initial conditions of Xi and s

(Meyer et aI., 1975; Miura et aI., 1980).
5.6.2.4 Predation (Predator- Prey Interactions)

A food chain of mixed populations indicated by the scheme (cf. Fig. 5.55)
(5.197)
contains bacteria as the first members in an aquatic and terrestrial environment,
while protozoa and ciliates represent other members of a biocoenosis. Thereby
a prey (xd is consumed by the predator (x 2 ).
Gause (1934) was the first to systematically study the interactions between
ciliates and their prey in a closed laboratory environment. The topic was
also studied extensively by Volterra (1931) who, together with Lotka (1925),
developed early mathematical models. The original Lotka - Volterra analysis
considered 111 and 112 constants, but normally they would depend on their
respective substrates. Thus
S1
111 = f(s) = I1max.1 K + (5.198)
1 S1
and
(5.199)
The model equations for simple prey-predator interactions, initially according

to Lotka and Volterra and later refined by Bungay and Bungay (1968), are as
follows:
(5.200)
and
(5.201)
where
k1 = killing efficiency constant based on encounters
k2 = predator growth constant, proportional to yield coefficient YX1/X2
k3 = specific dead rate

For a CSTR the model equations become (Curds, 1971)
(5.202)
rX2 = 112 X 2 - DX2 (5.203)

and
c. ,-------------------------,
AI X2
I,
I,
/,
"
,'
"
"
,1
/',\
,,
1
Xi'-"
_t
5.59 5.60
FIGURE 5.59. Computer simulation of the behavior of mixed populations (bacteria,

Xl' and ciliates, x 2 ) exhibiting predation on the basis of Equs. 5.202 to 5.204
(so = 200 mgl- 1 ; D = 0.1 h- 1 ; Jl.max.l = 0.6 h- 1 ; Jl.max,2 = 0.43 h- 1 ; KSI = 4 mg' 1-1;
KS2 = 12 mg '1- 1 ; YX11s = 0.45; Yx21s = 0.54). Limit cycle oscillations are the result.
(Reprinted with permission from Water Res., Vol. 5, Curds, Copyright 1971, Pergamon
Journals Ltd.)
FIGURE 5.60. Closed cycle relationship between predator x 2 and prey Xl from classical
theory following a phase plane analysis. For more information see Latka, 1925, 1956.
(See Equ. 5.205.)
Jl.l
I"s = -~-Xl + D(so - s) (5.204)
YX[S,
with Equ. 5.200 for the kinetics. A computer simulation of this set of predator
model equations is illustrated in Fig. 5.59 (Curds, 1971), which can also be
used for stability analysis. Three possible states may exist in this system,
depending on growth parameters: stable oscillations (limit cycles), damped
oscillations about the steady-state value, or an asymptotic approach to the
steady-state value.
If Equ. 5.200 is divided by Equ. 5.201 and integrated to Equ. 5.205
(5.205)
then each value of the integration constant gives a closed cycle for the relation-
ship between Xl and X z on an Xl - X z plane. This phase plane analysis, shown
in Fig. 5.60, demonstrates that the populations of Xl and X z oscillate around
a vortex point, the equilibrium point on the xdxz axis given by the ratio Jl.dkl
or k3/kZ (Canale, 1969). Biologists and also mathematicians have shown
considerable interest in the Lotka-Volterra model. Experimental data suggest
the existence of two singular points as derived by Sudo et al. (1975). Con-
sequently, the model shown by Equs. 5.200 and 5.201 with constant values of
k i cannot be fully adequate. Canale (1970) developed a modified predator-prey
model, indicated earlier with Equs. 5.198 and 5.199. Although Canale cited
the emergence of a limit cycle on the phase plane as responsible for the
oscillations, the experimental data cannot be attributed totally to the limit
t
23 4
/
~
/
,.-- _------------------5
I
'-------------------sin
FIGURE 5.61. Predicted behavior of a predator-prey system in single-stage CSTR

culture with respect to mean residence time t and concentration of limiting substrate
Sin for bacterial prey in the influent. A modification is represented by incorporating
a multiple saturation predator rate of Equ. 5.206 (after Jost et aI., 1973, and Tsuchiya
et aI., 1972). The curves 1-5 separate the following regions of the operating diagram:
(a) total washout, (b) X2 washed out, (c) no oscillations, (d) damped oscillations, (e))
periodic oscillations, and (ell) sustained oscillations.
cycle. In this situation Sudo et aI. (1975) proposed a more sophisticated model
by distinguishing between two parts of bacterial concentration: bacterial food
unavailable to the protozoa due to flocs and bacteria available because of
dispersion. Better agreement was found using a model incorporating microbial
floc formation and disintegration.
In the models, oscillations are inherent to the system. However, Tsuchiya
et aI. (1972) and Jost et aI. (1973) have demonstrated that under certain
conditions (combinations of dilution rate and S-concentration data) it is also
possible that predator-prey systems in a CSTR do not exhibit oscillatory
behavior. Results of stability analysis are shown in Fig. 5.61 (curves 1-4 with
regions a-e)) representing an operating diagram for the general dynamic
features for this case (after Jost et. aI., 1973). The diagram is based on Monod's
saturation model for growth of both Xl and X 2 (cf. Equ. 5.198) and shows
regions of different characteristics for given values of kinetic constants when
plotting liD versus So = Sjn' However, deviations from this model behavior
are known in experiments, leading to an important model incorporating the
multiple saturation predation rate of Equ. 5.206
J1.2 = J1.max.2(K 2 + x l )(K 3 + xd (5.206)
This equation reduces to the Monod-type Equ. 5.199 for Xl much larger than
K2 and K3 and has the meaning that J1.2 varies as xi when Xl is small. The
revised CSTR model comprising Equ. 5.206 can be analyzed according to
the operating diagram liD versus So. Figure 5.61 with curve 5 and regions
FIGURE 5.62. Mixed population

bioreactor problems illustrated
in a /1/s plot. (From Humphrey,
1977b.)
III
II
~ ____ ~~-L___________________ S
a through d and ell shows the behavior that is in accordance with the experi-
mental findings: that sustained oscillations disappear when t increases.
It is commonly accepted by ecologists that environmental heterogeneity
exerts a stabilizing influence on predator-prey interactions. This fact led to
investigations of combined wall growth in a CSTR with predation (Ratnam
et aI., 1982). Data obtained showed that bacteria but not protozoa were
attached strongly to the walls, and that wall growth had a significant effect
on the dynamics. A reasonable fit of all the experimental data was achieved
by combining a wall growth model (Topiwala and Hamer, 1971) with the
multiple saturation model for predator growth (cf. Equ. 5.206).
Beyond this field of fascinating possibilities for application and research
activities, some bioreactor operation problems should be referred to as typical
for mixed populations. An infinite number of stable steady states can appear,
for example, in the case of phenol degradation (Humphrey and Yang, 1975).
As shown in Fig. 5.62, the fermentation system contains two separate and
distinct species capable of metabolizing phenol. Species A is inhibited by
rather low levels of phenol, while species B is inhibited only at rather high
concentrations. The system behaves in a random way with respect to effiuent
phenol level and microbial flora. At operating point I in Fig. 5.62, the system
should settle on species B with s) in the effiuent. In the case of a low through-
put rate (point II), species A should dominate with low concentrations in
the effiuent (SII). Virtually any ratio of A to B is possible at the steady state
if operation is maintained at point III. Further, if fluctuations occur in the
influent, the effiuent concentration and cell mass relationship will wander all
over, depending in part on where they were last.
This phenomenon is thought to be much more general in biology than
expected (Humphrey, 1977a). It occurs with a single species growing on two
substrates when one metabolic pathway is repressed by the substrate of the
other. Another case is when a single species grows on a single substrate but
has two enzyme systems for handling the primary metabolite, one constitutive
and the other inducible.
These infinitely variable problems in bioreactor operation are waiting to be

solved. There have been relatively few systematic tests of the principles of
interaction of mixed populations. Experimental work on the basis of the
understanding from kinetic model theories should be encouraged by studies
utilizing a CSTR, a CSTR cascade, or a CPFR.
5.7 Dynamic Models for Transient Operation

Techniques (Nonstationary Kinetics)
Transient reactor operation plays an increasingly important role in bio-
processing and has to some extent already been considered (classification, see
Fig. 3.31; fed-batch culture, see Fig. 3.37; situation, see Fig. 4.4; guidelines to
solution, see Sect. 4.2 and Fig. 4.5; structured cell model concept, see Fig. 4.7;
application, see Chap. 6). Both "balanced" and "frozen" conditions have also
been considered in Fig. 3.34. A biosystem is in balanced condition when the
mechanism is fully adapted, as in a quasi-steady-state (if't'e 't'r)' All different
equations can be reduced to algebraic equations. A biosystem is in frozen
condition of the initial state (if't'e 't'r) and the mechanism may be neglected
due to the fact that the slowest step is rate determining ("rds concept"). By
this procedure, equations are reduced to parameters so that the number of
equations is reduced (e.g., the case of dropwise addition of substrate). This is
the case of steady state CSTR.
Intermediate periodic operation lies between both extremes and covers
systems in which the response time is of the same order as the imposed function
cycle time so that "resonance" is possible. Here the state of the mechanism
changes dynamically and is not directly related to the environmental condi-
tions at the moment considered. Intermediate periodic operations encompass:
semibatch periodic operations, selectivity in periodically forced CSTRs, cycled
reactors with heterogeneous catalysts, and non-steady-state biofilm reactors
(e.g., Rittmann and McCarty, 1981). Lag, overshoot phenomena, and oscilla-
tions can typically occur. Biological reactor operation will basically fall into
this latter class of operation, and balanced growth must be distinguished from
steady-state growth (Barford et aI., 1982).
5.7.1 DEFINITIONS OF BALANCED GROWTH AND

STEADy-STATE GROWTH
5.7.1.1 Balanced Growth

The concept of balanced growth was introduced by Campbell (1957) to describe
a metabolic state of a culture in terms of the distributed concentration (Xi) or
total mass (Xi) of a metabolic variable. According to Campbell's definition,
growth is balanced when the specific rate of change of all such variables is
constant. That is,
5.7 Dynamic Models for Transient Operation Techniques 273
1 dX i 1 dX i
-'-- = - ' - = constant (S.207a)
Xi dt Xi dt
This definition applies to a culture of microorganisms. In the case of an
individual microorganism, a broader definition is required (Campbell, 19S7).
1 Llx 1 LlX-
-_._' = -_._-' = constant (S.207b)
xi,ave Llt Xi,ave Llt
where
t = n' tg
n = any integer
tg = generation time [h]
t g = log21~
In ad9ition to the growth of a single microorganism, many other microbial
growth phenomena require a broader definition of balanced growth. An
example of this is periodic behavior or sustained oscillations (e.g., Heinzle
et aI., 1982). If such sustained oscillations are to be regarded as balanced
growth, the definition of balanced growth may be broadened, again using
Equ. S.207b but with
(S.208)
where n is any integer and tp is the period of oscillation [h]. These con-
siderations are a simple example of the important concept of regime analysis
(cf. Fig. 4.S).
Consideration also needs to be given to the question of how closely balanced
growth can be approached under batch growth conditions. While the ideal
experimental batch curve is often referred to (Herbert, 1961), the number
of reported experimental examples of balanced growth in batch culture is
extremely small (Barford and Hall, 1979b). In most cases, the experimental
data are not sufficient to allow any conclusive judgment as to whether
the growth is balanced. Therefore all metabolic rates are to be considered
(~, qs, qo, qc, qp, qHv' ... ). While batch cultures that approach balanced growth
have been reported (Barford and Hall, 1979b), it is possible that perfectly
balanced growth can never be achieved in a batch experiment. This is not
because of the limited time scale, but because the composition of the extra-
cellular medium in a batch culture is constantly changing and could be
expected to induce corresponding changes in cell composition.
Several important concepts are illustrated by these observations. First, it
is essential that a range of metabolic variables be considered before any
well-based decision can be made as to whether balanced growth has been
achieved. In addition, and most important, some metabolic processes achieve
a balanced growth condition more rapidly than others (e.g., ~ and RQ before
qo and qd. Consequently, the variable that is used as the criterion for the
attainment of balanced growth must be carefully chosen and explicitly stated.
5.7.1.2 Steady-State Growth

It is possible to consider steady-state growth as an extension of balanced
growth. That is, in addition to the definitions outlined previously, a further
requirement is that the overall time rate of change of the distributed con-
centration (xJ or total mass (XJ of any metabolic variable be zero. That is,
dx dX.
_1-_1-0 (5.209)
dt - dt -
Consequently, while balanced growth is a necessary condition for steady-state

growth, it is not a sufficient condition. Likewise, steady state is a sufficient
condition for balanced growth but is not a necessary one.
A final consideration from the definition of steady state is that no reference
to time scale is included. How long must a culture exhibit no time rate of
change of metabolic variables to be considered at steady state? Clearly this
should be related to the number of generation times without significant
variation in the magnitude of major metabolic variables, but how many
generation times and which metabolic variables? While tacitly it has been
generally accepted that 15 to 20 generation times would be excessive for the
achievement of steady state, many examples of much longer periods of time
necessary for steady states are available (Heineken and O'Connor, 1972).
Furthermore, very few experimental studies precisely state what criteria were
used in the determination of steady state (see Barford et aI., 1982).
5.7.2 MATHEMATICAL MODELING OF DYNAMIC

PROCESS KINETICS
As previously outlined, mathematical modeling uses both the methodological

concepts of unstructured and structured models.
5.7.2.1 Unstructured Models

In general, unstructured models can be considered a good approximation in
two distinct cases (relaxed steady state and quasi-steady periodic operation,
as shown in Fig. 3.34). However, unstructured models can successfully be
applied to batch processing with lags (e.g., Knorre, 1976) and to the dynamics
of stirred tanks (Agrawal et aI., 1982) by introducing a formal kinetic approach
for the lag time tL (see Equ. 5.72) and the variation in yield coefficient (cf.
Equ. 2.41). The number of model parameters is increased by t L , but the number
of process variables is only 3(S, X, t). Unstructured but modified modeling of
process kinetics has been successfully applied in the case of biological waste
water treatment, where nonstationary behavior occurs as a consequence of
fluctuations in feed stream or shock loading. Mona et aI. (1979) modified
Monod-type kinetics and adapted them to real situations by creating load-
dependent kinetic parameters (qS.max and Ks):
qS,max = kl . Bx (5.210)
and
Ks = k 2 Bx (5.211)
where Bx is the mass loading rate per unit biomass (sludge)
B = F's o [M.M-lT- l ] (5.212)

x V'x I X
Further, these authors introduced time lags 'LLl and 'LL2' responsible for the
time dependence
A dqs, max k B (5.213)

':IS,max + 'LLl ~ = l' x
and
-b dK s
1\.s + 'LL2Tt = k 2 Bx (5.214)
with qs,max and Ks being the deviation variables defined as

qs,max = qs,max(t) - qs,max (5.215)
and
(5.216)
The basis of this formal kinetic modeling is shown in Fig. 5.63 together with
the definitions of'LLl and qs,max'
A formal kinetic approach for product formation in transient conditions
STEP PERTURBATION FUNCTION

I
r----- ---
63 % 1. ....... .
I
I
I
I
Cis,max
--t
FIGURE 5.63. The time-dependent change in the maximum rate of substrate use, qs,maX'
as an illustration of nonstationary behavior (dynamic kinetics) of a bioprocess such as
this example of waste water purification: (1) easily metabolizable, synthetic medium,
(2) difficult to metabolize medium such as starch, (3) mixture of easily metabolized
substrate and suspended particles in waste water. 't"Ll = delay time for qS.max' (Adapted
from Mona et ai., 1978.)
was presented by Gutke et al. (1980)

kp = kp(fl, it) (5.217)
As the physiological state can be characterized in the most simple way by the
specific growth rate f1, the nonstationary state is quantified by the time
derivation of specific growth rate (it), being constant (ell) a given time period
during a shift-down
it = -ell (5.218)
with
f1(t) = f10 - ell' t (5.219)
The behavior of this model is illustrated in the computer simulation in
Fig. 5.64.
Another attempt at generalized rate equations was realized by Harder and
Roels (1981), applying the characteristic-time concept (cf. Sect. 4.2). A model
describing the dynamics of product formation was derived using Equ. 5.117
together with Monod-type kinetics for rs , incorporating endogenous metabo-
lism with the aid of Equ. 5.80, resulting in an expression for steady state
Q
qp = y:- f1 + Q. ms (5.220)
XIS
with Q]Jeing the product formation activity function, which is related in steady
state (Q) to f1
In xl Xo
-- ...... ,----------
\1
/1
I
I
I'II X
/ r- .~----------
I II
/
!I
-t
FIGURE 5.64. Kinetics of biomass concentration, logarithmic biomass concentration,
and specific growth rate according to Equs. 5.218 and 5.219 (solid lines; Xo = 0.2 gjl,
flo = 0.4 h- 1 , c = 9.1 h- Z ) and for comparison according to the well-known Monod
model (dashed lines; Xo = 0.2 gil, flmax = 0.8 h- 1, Ks = om gil, So = 0.49 gjl, YxlS = 0.5).
(From Gutke et aI., 1980.)
Q = f(fl) (5.221a)
or
qQ = J1" Q (5.221b)
From Equ. 5.220 and 5.221a it follows that
Q= Yx1s ' qp,max

(5.222)
fl + ms' Yx1s
The extension to dynamic situations is achieved by the formulation of a rate
of adaptation of Q to environmental conditions with the aid of the intrinsic
balance equation
(5.223)
with rQ the rate of synthesis of Q. The difference between actual rate of Q

synthesis and the steady-state rate is assumed to follow
(5.224a)
or
(5.224b)
where Kadapt is a positive constant (constant of adaptation) given by
of
Kadapt = - oQ _ (5.225)
Q=Q
This constant can be estimated from shift-down or shift-up experiments.

Finally, combining Equs. 5.221b, 5.223, and 5.224b for the rate of change
of Q, one has the result
Q= -(KadaPt + fl)(Q - Q) (5.226)
The relaxation time tcharact for the adaptation to a new steady state during
a shift in continuous culture is given by
1 1
tcharact = (fl + K adapt ) = fl (1 + K adapt /) (5.227)
fl
If Kadapt is large, a new steady state will be reached almost instantaneously
with no lag of the organism. If Kadapt is small, the time constant for adapta-
tion will be equal to 1/fl, that is, dilution through growth will control the
adaptational process.
Figure 5.65 represents the apparent relationship between Q and fl according
to Equ. 5.226 in dynamic situations. As can be seen, when Kadapt is large the
steady-state relationship (see Equ. 5.116) is obtained.
Esener et al. (1981) applied unstructured modeling to fed-batch processing
Cj
A
qp
0.5
EO 5-116
f(Kadapt )
EO 5-226
0
0
iJ _t
5.65 5.66
FIGURE 5.65. Dynamic relationship between specific rate of product formation qp and
specific growth rate J1 for various rates ofthe exponential decrease of the specific growth
rate expressed with the aid of varied values of the constant Kadapt (see Equ. 5.226).
The steady-state relationship is obtained at large K values. (From Harder and Roels,
1982.)
FIGURE 5.66. Typical plot of growth kinetics and macromolecular composition of

microorganisms during the course of a batch fermentation according to Herbert (1961).
Concentrations of dry biomass (x), individual cell weight (w), number of cells per
volume (n), and RNA and DNA per unit weight.
on the basis of Equ. 3.92 for the concentration of X and S. If such an unstruc-
tured model is combined with the elemental and energy-balance principles
(see Sect. 2.4), the time dependence of O 2 consumption and CO 2 and Hv
production can be calculated with great success for batch, continuous, and
fed-batch cultivation. The results of this investigation showed that in those
cases where detailed kinetics playa role (i.e., in the shift from a process
controlled by the maximum specific growth rate of the organism to one where
growth is limited by the addition rate of the substrate), the model provides
only a poor fit to the phenomena actually observed. Structured models,
therefore, should be more adequate for the description of a system during
highly transient periods.
5.7.2.2 Structured Models

In structured model building one must select the parameters that are the most
relevant for the description of the physiological state of the organism. Infor-
mation from molecular biology, chemistry, microbiology, and biochemistry
is required for this purpose. Logical first choices would be the DNA, RNA,
carbohydrate, and protein contents of the cells for describing their physio-
logical state. All of these variables can be experimentally determined, and
their dependence on steady-state growth rate is well established, as shown in

Fig. 5.66 (Herbert, 1961). It must be noted, however, that even the consideration
of these four relevant components is not sufficient to describe the activities
and qualities of the organisms fully; for example, no information can be
obtained about the geometrical structure of the cells or the dependence of
diffusional processes on such structures (Kossen, 1979). Thus, a structured
model should be formed in such a way that it provides information only about
the most relevant processes and variables (see Sect. 2.3). For a comprehensive
description, too many parameters would need to be incorporated into the
model. Such complicated models (e.g., Fig. 2.22) are mathematically very
complex to manipulate, and most of the parameters often lose their biological
significance. For such models, numerous critical experiments have to be
performed to determine the parameters.
Simpler models can be Qbtained by considering a few variables as an exten-
sion of the unstructured models, which consider one biotic variable (Esener
et aI., 1982b). These models, in which the activity of biomass is specified by
more than one and up to three or four variables, are called "compartmental
models." They have moderate mathematical complexity and are easier to
verify experimentally. The first formal compartmental model was due to
Williams (1967,1975), who showed that even a simple model with two com-
partments could describe most of the experimentally observed phenomena,
at least qualitatively. Some inconsistencies of this model were later corrected
by Roels and Kossen (1978) and Roels (1983). Ramkrishna et al. (1967) and
Fredrickson et al. (1970) have contributed greatly to the theoretical aspects
of structured modeling.
The building of chemically structured models consists of (a) setting up of
mass balances for the relevant components of the biomaterial over an appro-
priately chosen boundary, (b) postulation of the relevant kinetic expressions
for reactions taking place within the biotic phase and its boundaries, and
(c) evaluation of the constraints imposed by elemental balances and the
application of thermodynamic principles. With reference to Fig. 5.66, it can
be concluded that RNA and carbohydrate fractions change most in response
to changes in steady-state growth rate. Thus, it might be desirable to consider
these components explicitly in a structured model. Protein content changes
only slightly over the same range. It is difficult to deal with carbohydrate as
a separate compartment: The absolute intrinsic concentration of carbohydrate
is very low (about 5% dry weight) and hence variations in it are difficult to
detect analytically. Therefore, in the model to be presented, RNA, carbo-
hydrate, and other small cellular molecules are lumped into one compartment,
which will be referred to as the "K compartment." Since RNA is the main
constitutent, the amount in the K compartment may be expected to be
proportional to the RNA content. The other compartment is called the G
compartment and contains the genetic material and all the rest of the cell
constituents, that is, proteins, DNA, structural material, and so on. This
approach more or less assumes that the RNA concentration (and thus the
FIGURE 5.67. Concentration/time plot of struc-

tured biomass model according to Williams
(1967): Biomass G and K compartment (x G , xd
and total biomass (x) and substrate (s) simulated
on the basis of Equs. 5.228d-f with kJ = 0.0125
and k2 = 0.025 .
..... ~..'. .. ~.K.
RNA synthesis) is the bottleneck. Since RNA plays a central role in the
synthesis of proteins, this approach to structured modeling seems reasonable.
A diagram of the two-compartment model was shown in Fig. 4.8.
The following reactions can be formulated, based essentially on the work
of Williams (1967) and on its modifications and extensions by Roels and
Kossen (1978) and Roels (1978):
G
S--+ YsKK rate'sK (5.228a)
K
K --+ YKGG rate 'KG (5.228b)
G--+K rate 'GK (5.228c)
The G compartment is assumed to be synthesized from the K compartment
under the catalytic action of the K compartment. The latter compartment is
thought to be synthesized from substrate under the catalytic action of the G
compartment. This is a great simplification of the complexity of the cellular
processes, but is significantly closer to reality than the unstructured approach.
A computer simulation of the Williams model is shown in Fig. 5.67, illustrating
the basic behavior of the two-compartment model (Williams, 1967).
The next step is the postulation of rate expressions for 'SK, 'KG' and 'GK'
For a description of the state of the culture as a function oftime, the approach
advocated by Harder and Roels (1982) can be used without problems when
the kinetics and the stoichiometry of the processes involved are defined:
1. Conversion of the substrate to the K compartment. (rate 'SK and stoichiome-
try YSK)' From Equ. 228a the following relationship is valid:
(5.229)
The rate of substrate consumption is assumed to be of saturation type,
that is,
kSK's wG
'SK = - - - ' 'x (5.230)
Ks + S KG + WG
Here kSK represents the maximum value 'SK can achieve and WG is the mass
fraction of component G.
2. Transformation of the K compartment into the G compartment (rate r KG

and stoichiometry YKd. The following relationship is proposed:
rKG = !3(WG )' x (5.231)
The value of rKG can be described by
rKG = kKG'WK'WG'x = k KG 'wd1- wK)x (5.232)
since W K + WG = 1, by definition. Here, the rate of K to G is assumed to be
influenced by the intrinsic concentrations of K and G as well as by the total
biomass concentration. An interesting observation here is that when G = 0,
Equ. 5.232 predicts no G formation. This has a biological explanation:
When G = 0, enzymes and DNA necessary for G synthesis are absent.
3. Turnover of the compartments of the biomass is modeled by
(5.233)
where mG is the specific turnover rate of compartment G (maintenance rate).
It is assumed to be a depolymerization process and is first order in the total
amount of G compartment (wGx), The specific rate of depolymerization is
mG'The yield constant for the formation of the K compartment from the G
compartment is assumed to be unity, that is, no mass is lost during the
depolymerization of the G compartment to precursors.
The balance equations for the rate of change of the substrate concentrations,
the biomass concentration, and the fraction of the G compartment are now
obtained by the application of the formalism treated by Harder and Roels
(1982). The resulting equations are
ds S WG
- - -q --'
dt - S.max Ks + S KG + WG x+FS (5.234)
with Fs the flow of substrate to the system [kg m -3. h -1],

dx S WG
-d = YsK' qS.max-K x + (YKG - 1)!3(W G )X + Fx (5.235)
t s +
S KG WG +
with Fx the net flow of unstructured biomass dry weight to the system
[kgm- 3 h- 1 ],and
dWG S WG
dt = - YsK' qS,max Ks + s KG + WG . WG + !3(WG )
X {wG + YKG (1 - wG)} - mGwG (5.236)
These equations contain the transport contributions Fs and Fx , which depend
on the mode of operation and are summarized in Table 5.5. A most important
feature of Equs. 5.234 through 5.236 is the fact that the intrinsic balance
equation is independent of the mode of operation.
The necessity for a model of such simplicity, while sufficient knowledge is
available for the construction of a model of much greater realism, may not be
TABLE 5.5. State equations of the Williams two-compartment model

according to Harder and Roels (1982).
For batch operations 0 0

For CSTR with feedback of fraction (1 - wD ) of D(s;n - sex) -WD ' D X
biomass leaving system
For fed-batch culture Fs(t) 0
FIGURE 5.68. Block diagram showing a

simple model of metabolism; I, inter-
mediary product or precursor. (From
Roels and Kossen, 1978.)
obvious. Two factors should be considered:

1. A number of regulatory mechanisms at the level of energy generation and
consumption operate with such small relaxation times that a pseudo-steady-
state hypothesis with respect to these mechanisms is justified. Hence, the
introduction of these details seems to be unnecessary.
2. A minimum of complexity is desirable because the complex model often
proves very difficult to verify and may fit experimental results without
having any relationship with the behavior of the organism. Roels and
Kossen (1978) gave a block diagram of a simple approach to metabolism,
shown in Fig. 5.68. Only after obtaining experimental evidence that the
simple model should be rejected because of insufficient fit of the data or
unrealistic parameter values should additional complexity be introduced.
If the relations obtained for rs are compared, the structured model can be
seen to have a maintenance term dependent on the growth rate in the form
of the following equation:
kKG D + mG
ms = y-. (k . Y. )2 [kKG YKG - (D + mG)](l - YKG) (5.237)
.18K KG KG
The two-compartment model exhibits many features observed in batch and
continuous culture experiments. The method is certainly promising as an
adequate approach to transients.
5.8 Kinetic Models of Heterogeneous Bioprocesses 283
The particular model presented, however, must be considered as a pre-

liminary proposal because many of the kinetic assumptions do not rest on
solid biochemical facts about the internal regulation of the cell. Furthermore,
there are difficulties in identifying the compositional nature of the K and
G compartments in terms of structural components of the cell. It is clear that
a more thorough study of known regulation phenomena and an empirical
study of transient situations, for example in continuous culture, is needed,
especially in the CSTR.
5.8 Kinetic Models of Heterogeneous Bioprocesses

5.8.1 BIOFILM KINETICS
While pseudohomogeneous rates of biokinetics follow the equation for rj
[kg'm- 3 'h- 1 ]
(5.238)
with kl and k2 defined in Equ. 4.25, heterogeneous rates are expressed as
rates r; [kg m- 2 . h- 1 ] related to the area of active biomass as [m 2 . m- 3 ]
(5.239)
The parameters k3 and dp are responsible for heterogeneity, dp being the
particle diameter and k3 a process kinetic constant, defined by Atkinson
(1974) as
(5.240)
with k~ = as' k 1 . The parameters kl' k2' and k3 are referred to as coefficients
of the "biological rate equation" (BRE) (Atkinson, 1974). Parameter k3 repre-
sents the interactions between reaction (k 1 ) and physical internal transports
(Ds.err , related to dp by the two-film theory given in Equ. 3.30a).
The BRE (cf. Equ. 5.239) is, valid for heterogeneous microbiological and
biochemical systems, includilll fermentations and enzyme engineering systems.
The model uses the same type of equation as Equ. 4.60 with limiting boundary
conditions Equ. 4.61a and b, with the exception that the substrate utilization
rate is referred to the surface area ofthe biological mass, r~, which is a quantity
more easily measured. The connection with the specific rate, qs [hr- 1 ], is
through the relationship
(5.241)
where as is the specific surface area of the biological mass, which has a density
p. A graphical representation of the BRE is depicted in Fig. 5.69, plotting
r~ versus s in analogy to the Monod-type diagram (cf. Fig. 5.13). Figure 5.69
can be applied directly to biofilm kinetics (r~).
qs,max ...
d / " I ""
~ I
I --,/, /'
,---,-- Monod - -/_-/
1-,"
----T--
I
1/2 ......... " --~d
,:
I /
I
'Monod
>L-----------_s
5.69 5.70
FIGURE 5.69. Biological film kinetics: Plot of the rate of substrate use, r~, referred to
the effective biological film surface area, as a function of the substrate concentration, s,
at various thicknesses of a biological film, d. A realistic limiting case (--) is found at
low s where formal first-order kinetics in s are found. (Adapted from Atkinson, 1974;
with permission of Pion, London.)
FIGURE 5.70. Kinetics of heterogeneous bioprocess: Diagram of the specific rate of

substrate use, qs, as a function of the substrate concentration, s, varying the thickness
of the biocatalytically active mass, d, for floc and film geometries. Monod kinetics
appear as the boundary case at low d. Unrealistic cases are indicated with dashed lines.
(Adapted from Atkinson, 1974; with permission of Pion, London.)
The complete computer simulation of the BRE is given by dotted lines

for various film thicknesses d, and the area of experimental verification is
indicated with solid lines. Obviously, high S concentration cannot be attained
with thick films. As can be seen from Fig. 5.69, the appearance of a simple
first-order reaction rate in cases ofbiofilm processing can often be understood
as an observed case of pseudokinetics.
With biological floes, reaction rates should be plotted as illustrated in
Fig. 5.70 (Atkinson, 1974). With increasing film thickness, and with larger
particles, the curve moves to higher S values. Again, dotted lines represent the
computer simulation, while solid lines correspond to experimental situations.
The BRE shows that only one kinetic equation can exist for various ge-
ometries of the biological mass. In the past, attempts were often made to
formally utilize zero-order equations with microbial particles (floes) and to
formally use first-order equations with microbial films, and this discrepancy
was the starting point of Atkinson's work.
For both types, one may recognize Monod-type relationships that omit
Ds. For biological films, one actually has a realistic zero-order equation
(Equ. 5.244). Higher substrate concentrations are attainable with floes, so that
here zero order also dominates.
In Fig. 5.71 various limiting cases for microbial particles (floes) are presented
/FLOC~
d p - - small dp - high
(homogeneous) (heterogeneous)
! /
c - - small
l
~
c - - high
~
k1 "a k1
r= - - ' c r = r max = k2 P
k3 "P
(5.242) (5.243) (5.244)
FIGURE 5.71. Equations derived from the biological rate equation (Atkinson, 1974) that
can be used for the quantification of different cases of "biofloc processing."
/FlLM~
dp -- small d p __ high
-----
(pseudohomogeneous) (heterogeneous)
! /
c __ small
!
c -high
k2_'c_ , ,
~ k1
r'=r' __
max 1 + k c
r = rmax = k d
2 2
(5.245) (5.246) (5.247)

FIGURE 5.72. Equations derived from the biological rate equation (Atkinson, 1974) that
can be used for the quantification of different cases of "biofilm processing."
along with Equs. 5.242 through 5.244 derived from the BRE. The same are
shown in Fig. 5.72 for the thin-film geometry (Equs. 5.245-5.247).
The BRE in the form of Equ. 5.239 can be rewritten to give
(5.248)
equivalent to
(5.249)
using the Thiele modulus rP (cf. Equs. 4.74 and 4.80) instead of Equ. 5.240
(see also Sect. 4.5.2). The consequence of this type of modeling is that the
complexity of interactions between biological reactions (k r ) and physical
transports (k TR ) can be significantly reduced with the aid of the concept of
efficiency '1r interpreted in Equ. 4.53 as a function of rP or Dan (cf. Fig. 4.30).
With the '1 concept, the truly heterogeneous nature ofbioprocessing is simplified
to the pseudohomogeneous case, where the algebraic factor '1r is only a multi-
plicative term in the pseudohomogeneous rate. Thus, for Monod-type kinetics
we have
S
rj
.
eff = rj .m aKsx -+- -S ' X '11r (5.250)
In the literature are many articles on porous diffusion, especially in con-

nection with carrier-bound enzymes or cells (for example, Pitcher, 1978).
These are directly connected to the principles expressed in Sect. 4.5 concerning
the influence of internal and external mass transport. The results are presented
in the same graphical form as Fig. 4.36 in which the effectiveness factor of the
reaction '1r is presented as a function of the Thiele modulus. For formulating
an appropriate moduls one needs knowledge of the difficult to measure Deff
value. The following equation has shown itself useful in that the volume-based
reaction rate rs is obtainable directly from the experimental measurements
(Pitcher, 1978) (cf. Equ. 4.74)
(5.251)
with rS.eff the observed effective rate per unit volume of porous carrier
[kg m -3. h -1] (Pitcher, 1975). Since the effectiveness factors for intermediate
Ks/s values lie between the curves for the limiting cases of n = 0 and n = 1,
llr
0,8
0,6
0,4
0,2
q2 P
FIGURE 5.73. Plot of the effectiveness factor of reaction, I)" as a function of the modified
Thiele modulus, cI> (Equ. 5.251) for enzymatic reactions. Variation of the Ks value are
shown in the case of a flat-plate geometry of the solid phase (Pitcher, 1978). This plot
is analogous to Fig. 4.36.
we see from Fig. 5.73 that I1r is indeed relatively insensitive to the remaining
intrinsic parameters Ks/s and k 2 Figure 5.73 is an equivalent form to demon-
strate the BRE, as shown in Figs. 5.69 and 5.70. Variations of the plot in
Fig. 5.73 are described in the literature with modified definitions of a modulus
(e.g., Shieh, 1981) or change in the functional dependence, as shown in Fig. 5.70.
Numerical solutions are recommended (Moo-Young and Kobayashi, 1972)
for cases involving complex kinetic equations such as substrate or product
inhibition.
In addition to the process kinetic formulations of Atkinson, which are derived
from the theory of mass transport with simultaneous biological reaction,
other kinetic equations, macrokinetic in origin, are also known.
With reliance on the external, apparent nature of the influence of mass
transport, as shown in Fig. 5.70, where with an increasingly thick biological
mass there is a higher Ks value, an equation based on Monod kinetics has
been proposed (Powell, 1967), already shown in Equ. 5.43, where
(5.252)
Keff is the sum of the Monod constant Ks and a formal kinetic coefficient for
the volumetric mass transfer resistance, K D , in the solid phase. The behavior
of this approach at varied ratio KD/Ks is shown in Fig. 5.74. The application
of this equation has been encouraged by Humphrey (1978).
A second alternative for a kinetic equation has been proposed by Harremoes
(1977, 1978): It is a kinetic description of the plot in Fig. 5.70, and is an
equation formally one-half order in S (cf. Equ. 4.87)
(5.253)
The integrated form of such an equation has already been discussed (Equ. 5.35).
To illustrate the range of applicability of each of these approaches (n = 0,1)
and the transitions between the cases, it is convenient to use dimensionless
quantities A, B, E, which are simply the ratios between the three different
bulk reaction rates, interrelated as A 2 = E B.
KD/KS
r------ 00
10
2
o
FIGURE 5.74. Normalized kinetic
plot of the Powell approach to
internal transport limitation (cf.
Equ. 5.43) with variation of the
ratio KD/Ks. Monod-type kinet-
ics appear when this ratio is zero
(after Humphrey, 1978).
O~/------~----~2--------A
FIGURE 5.75. Dimensionless plot of reaction rate K = r'/k'o versus substrate concentra-
tion in the bulk liquid A (cf. Equ. 5.254a). Zero-, first-, and half-order reactions
(ns = 0,1, 1/2) are demonstrated in their range of applicability. Note how ns = 1/2
may erroneously be interpreted as the saturation effect of Monod kinetics. Band E
are explained in Equs. 5.254b and c (Harremoes, 1977).
k~
A=-s (5.254a)
k~
B=kI/2 s (5.254b)
k~
kI
2 's
E=-k (5.254c)
1/2
Figure 5.75 shows the bulk reaction rate of the biofIlm K in dimensionless form
as a function of the bulk concentration A with different biofilm characteristic
values of AlB (Harremoes, 1978). Even though the depth of penetration may
be small compared with the entire film thickness, the effect of internal diffusion
resistances is to mask the true zero-order kinetics (see Equ. 5.111), yielding
an apparent reaction order of one-half.
Formulations involving simultaneous internal and external limitations are
to be found in the literature (Frouws, Vellanqa, and Dewilt, 1976; Hamilton,
Gardner, and Colton, 1974) and are referenced in Sect. 4.5.3.
5.8.2 UNSTRUCTURED MODELS OF PELLET GROWTH

Fungal growth occurs in two morphologically distinct forms: mycels (fIlamen-
tous hyphae dispersed in fluid) and pellets (stable, spherical agglomerations).
In submerged cultures, growth in fIlamentous form follows the exponential
law (Metz and Kossen, 1977; van Suijdam et at, 1982). Individual hyphae
grow only at the tips (see Equ. 5.109) and have a linear extension rate.
Exponential growth is maintained by continuous branching of the hyphae.
This can be expressed by the "hyphal growth unit," the mean length of hyphae
per growing tip. The combination of a linear extension rate with a constant
hypha.! growth unit concept results in exponential growth, where the unit is
independent of Il (van Suijdam and Metz, 1981).
The growth of pellets is not always exponential because of mass transfer

limitations (Metz and Kossen, 1977; Pirt, 1975; Whitaker and Long, 1973).
Increasing the size of pellets results in a significant decrease in reaction rate,
and the apparent Ks value increases (e.g., Kobayashi et aI., 1973). To describe
the growth of pellets, one can use either the exponential law or the Monod
relationship. However, the logistic equation (Kendall, 1949) in the form of
Equ. 5.74 or the Gompertz equation (Chiu and Zajic, 1976)
"x = k j ' X . exp( - k j t) (5.255)
is often preferred. In Equ. 5.255, k j and k j are constants with the dimension
[h -1].
Most often, the model used to describe the growth of pellets is the "cube root
equation" (cf. Metz and Kossen, 1977; Pirt, 1975; Trinci, 1970). Considering
the cell mass to be related. to the pellet radius, R, and the number of pellets,
N, according to Equ. 5.256
N 4n 3
X=-p-R (5.256)
V 3
the cube root law is often expressed in two different ways:
X I1/3 = (X1XJ/3. t (5.257a)
X{/3 = XJ /3 + k t (5.257b)
with (Xl a proportionality constant and with
(5.258)
where dp is the thickness of a peripheral zone around the pellet; the zone
alone contributes to growth (Pirt, 1975).
The appropriate rate equation for this law is
(5.259)
where (X2 [kg 1/3 . m -2J is a constant.
The cube root law results from a combination of exponential growth with
mass transport limitations in the solid phase, which is expressed in the concept
of dp as shown by Pirt (1975).
The predictions of different kinetic model equations for pellet growth are
compared in Fig. 5.76 (van Suijdam et aI., 1982). The main objection against
the models presented is that they are autonomous, that is, the biomass rate
equation is not related to the concentration of the limiting substrate, and that
mass transfer limitations are not explicitly considered in these macrokinetics.
To introduce these phenomena, a combination of mass transfer limitation
effects and biokinetics must be used. Such an integrated model for pellet
growth was developed (Metz, 1975) and extended (van Suijdam et aI., 1982),
based on balance equations for X, S, and O 2 in liquid and solid phase.
FIGURE 5.76. Comparison of various rate equations in integrated form in a normalized

plot of biomass concentration x/x o versus time t = J-lma.t in case of pellet growth:
exponential law rx = J-lma.X (--); Monod equation, see Equ. 5.38 (---); logistic law,
see Equ. 5.75 (------); Gompertz' law, Equ. 5.255 (-----); cube-root law, Equ. 5.257
and 5.259. Parameter values xmax/xo = 250, k; = 5.5, C( 2 dp = 1.5 kg 1/3 . m- 1 (van
Suijdam et aI., 1982).
5.8.3 LINEAR GROWTH

The characteristics of a linear growth pattern as the consequence of an enzyme
system with constant activity have been thoroughly investigated and modeled
by Knorre et al. (1978a,b). It is thought that linearity is caused by constant
activity of enzymes in systems where substrate concentration is limited.
Linearities always indicate the presence of some limitations; even their exact
nature cannot be readily determined. Beyond a lack in nutrients, linear growth
phases can also be the result of transport limitations. Thus, from a systematic
point of view, these data repesent pseudokinetics.
5.9 Pseudokinetics
Pseudokinetic parameters result from a kinetic study whenever the model
(either knowingly or unknowingly) is simpler than the real situation. Generally,
falsification can be the result of the undetected influence of other reactions
or of transport phenomena (macro kinetics).
Pseudokinetic phenomena become evident only when process kinetic anal-
ysis is carried out with mathematical models. Most bioprocesses are basically
heterogeneous systems. Generally, pseudohomogeneous rates measured in
L phase analyses are used, because they are thought to reflect directly the
intrinsic reaction rate of metabolism in the solid phase (biomass). Even under
steady-state conditions, however, this assumption is not necessarily valid,
5.9 Pseudo kinetics 291
because different concentration levels may create different reaction rates.

Barford and Hall (1979a) demonstrated that external overall flux measured in
the L phase does not even approximately reflect the true internal fluxes of
metabolism. This fact suggests caution in accepting any model as a "true"
reflection of the in vivo state.
Many factors are known to influence the shape of Monod-type biokinetics
(A. Moser, 1981; Fiechter, 1982); most of them alter Ks. Apparent Ks values
can be the result of:
- Multi-S limitation
- Insufficient mixing in the liquid phase
- External transport limitation
- Internal transport limitation
- Ionic strength (Hornby et aI., 1968)
- High cell concentration (Contois kinetics)
- Endogenous metabolism
- Non-stationary processing
- Product inhibition
- Biosorption
One might also think that the Ks value depends on the size of the cells
(transport limitation by Ds ), and that only in some ideal cases is Ks the same
as that appropriate to the enzyme or mitochondria (Hartmeier, Bronn, and
Dellweg, 1971; Kessick, 1974).
In case of 2-S limitation, the apparent Ks value is given by
KS2
K s, Lapp = KS1 ( 1 + ----;; + KS2 . 8821 ' KS1 ) (5.260)
Analogous cases of kinetic equations for heterogeneous situations show similar

behavior, explaining apparent Ks values by external mass transfer limitation:
Hornby et ai. (1968)
r:nax
Ks,app = Ks + -k (5.261)
L2
where r;"ax is the maximum reaction rate per unit external surface area
Schuler et ai. (1972)
K - K [r;"ax . Ks ] (5.262)
S.app - S + kL2(8 o + Ks)
Kobayashi and Moo-Young (1971)

3 r;"ax
Ks,app = Ks + ~k (5.263)
L2
Finally, a case of pseudokinetics appears in case of global measurement

techniques, as in biological waste water treatment.
The BOD, COD, and TOC are widely used. These values, however, are not
correlated strictly (e.g., Aziz and Telbut, 1980). As a specific example the BODs
value should be mentioned. This measurement (in a time interval of 5 days)
reflects not only the O 2 used for metabolizing the primary substrate, as
illustrated in Fig. 5.77, but also other O 2 demands (for endogenous metabo-
lism, for other strains in the food chain of biocoenosis, cf. Fig. 5.55, and for
metabolism of N-containing substances). As indicated in Fig. 5.77, only the
BODPL (after 1 day) is a true value for S consumption and thus is the reliable
value for design purposes (Wilderer et aI., 1977):
(5.264a)
or
qo = Yo1s'qs (5.264b)
where the yield coefficient Yo1s is called "specific O2 utilization."
5.10 Kinetics of Sterilization

Techniques such as sterilization, pasteurization, or disinfection operate on
the basis of the effects of heat, radiation, or chemical agents on viability of
cells (bacteria, viruses), while filtration and membrane techniques act mainly
mechanically. Even though the mechanisms of desactivation are quite different,
similar basic concepts can be applied in most of these cases.
5.10.1 BASIC KINETIC ApPROACHES IN STERILIZATION KINETICS

The death rate for a population of microbial cells is commonly found to obey
a logarithmic law, and this is expressed by formal first-order kinetics
dN
-Tt= kd'N (5.265)
where N is the concentration of viable cells [number/ml] and kd is the specific

death rate constant [h -1 ].
Figure 5.78 is a simple plot of cell number versus time at a given temperature
according to the logarithmic death rate. To achieve a certain level of sterility
(Nst ) it is clear that a time tst is needed. A normal value for Nst is about 10- 4
cells/mI.
A consequence of the first-order law is that t --. 00 for Nst = O. However,
the first-order concept cannot predict the behavior of a single cell; it is only a
statistical approach to the behavior of a large population. Integrating Equ.
5.265 with initial conditions N = No and t = to gives
(5.266)
5.10 Kinetics of Sterilization 293
L
N
1
9""J.1qd NI NO
...::'1---1
n =1 ","
tot 'V" qolXjl i
1
_-----
----II BODS
T=const.
...... 1
~~~~ I
qd SI :t
IBODpL Nst -t
t
5 [d] tst
5.77 5.78
FIGURE 5.77. Schematic diagram of the time course of biological oxygen demand
represented in a graph of oxygen concentration 0L versus time, t, in days, showing
a succession of different steps: O 2 consumption due to waste degradation (qo{s});
maintenance requirement (mo); O 2 consumption by the "food-chain" organisms Xi
(qo {xJ); O 2 utilization due to nitrogen removal at high residence time (qo {N}).
Interpretation of BOD5 values thus shows that these values are of low significance for
engineering design purposes. As a substitute, a "plateau BOD" (BOD!) was proposed
to occur after one day (Wilderer et aI., 1977).
FIGURE 5.78. Plot of the decrease of cell number N due to sterilization effects in
time t at a given and constant temperature. The "sterilization niveau," Nst (normally
10- 4 cellsjl) is reached at t st ' the time needed for sterilization.
which can be plotted to give the linear form shown in Fig. 5.79. Death rate
constant kd is a function only of temperature. Deviations from the simple
logarithmic death rate are observed, especially with bacterial spores, as il-
lustrated in Fig. 5.80. Another case is included in the figure, where some
activation by temperature occurs before the number of spores decreases. For
the inactivation of spores, a death model was proposed by Prokop and
Humphrey (1970) in which a resistant type of spore Nr proceeds to death (Nd)
via a sensitive intermediate state Ns . Thus
N~N~N
r n=l S n=l d (5.267)
The corresponding differential equations are
dNr
-d = -kd ' l'Nr (5.268)
t
and
(5.269)
The solution takes the form

InN
InN ,-.
! '.,
InN./ \T'ACTIVATION
o, \
",
,, \.
\
, ,,
, SPORES
"-
CELLS
~----------------t
5.79 5.80
FIGURE 5.79. Typical plot of sterilization kinetics with viable cells, according to
Equ. 5.265 with varying temperature. Death rate constants kd can be estimated from
the slope of the lines.
FIGl]RE 5.80. Typical plot of sterilization kinetics with bacterial spores compared with
vegetative cells. A deviation is known due to the activation of spores by temperature.
N
N,= k k~lk [ exp(kd. 2 t)- kkd.2exp(-kd.lt) ] (5.270)
o d,l d.2 d,l
Alternative expressions for the kinetic constants kd are sometimes used in
different technologies (see Equs. 5.8 and 5.9). Supplementary to the "D value"
and "Z value," an "F value" is known, representing the value of tst at 121C.
Another case oflogarithmic behavior appears when air is sterilized (filtered)
by fibrous media, and the "log-penetration law" is valid. This law relates filter
effectiveness (N/No) to the filter thickness L and a factor KF
N
In-- = -KF'L (5.271)
No
The filter constant KF depends on fiber diameter, volume fraction of fibers in
the filter, and a fiber effectiveness factor (Chen, 1955).
5.10.2 MULTICOMPONENT SYSTEMS IN FOOD TECHNOLOGY

Almost all reactions taking place in foodstuffs, such as quality loss due to
the loss of vitamins, enzymes, or color (chlorophyll); to the death ofliving cells
or microbial spores; or to other chemical reactions, follow formal first-order
kinetics. Reactions associated with browning are often described by zero-
order kinetic reactions (Saguy and Karel, 1980).
Thus, the technical handling problems can schematically be described with
the aid of Fig. 5.81, where a series of components exhibit different k values.
As indication for sterility in food technology, the heat-resistant enzyme per-
T= constant
c end 7
~--~-----------+I------------------
tend = tst
FIGURE 5.81. Schematic representation of a bioprocess in a multicomponent food-

processing system in a normalized concentration/time diagram. The kinetic rate con-
stant, k, of reactions 1-7 is (A. Moser, et al. 1980b):
k [S-1 ]
1. Microbial cells 10 1 _ 10 13
2. Enzymes 10 1 _ 1013
3. Spores 0.5-20
4. Thermoresistant enzymes (peroxidase) 0.37
5. Chlorophyll 0.2
6. Vitamins 0.02
7. Nonenzymatic browning (Meillard) reactions 0.008
oxidase is often used with k = 0.37 S-1 at T = 121 DC or the spores of Bacillus
stearothermophilus (k = 2.9 min- 1 at T = 121 DC) are used for fermentations.
To obtain sterile foodstuffs that can be stored for long periods of time,
cells, spores, and enzymes must be destroyed while at the same time vitamine
content, color, and aroma are not substantially affected. This is possible using
kinetic data and a reactor with an average processing time t = t St ' Since a first-
order reaction is involved, the optimal reactor configuration is a continuous
tube reactor (Moser, Kosaric, and Margaritis, 1980b).
BIBLIOGRAPHY
Adair, G.S. (1925). J. BioI. Chem., 63, 529.

Agrawal, P., Lee, c., Lim, H.C., and Ramkrishna, D. (1982). Chem. Eng. Sci., 37,453.
Agrawal, P., Lim, H.C. (1984). Adv. Biochem. Eng. Biotechnol., 30, 61.
Aiba, S. (1982). Adv. Biochem. Eng., 23, 85.
Aiba, S., and Hara, H. (1965). J. Gen. Microbiol., 11,41.
Aiba, S., Humphrey, A.E., and Millis, N.F. (1973). Biochemical Engineering. New York:
Academic Press.
Aiba, S., and Shoda, M. (1969). J. Ferment. Technol., 47, 790.

Aiba, S., et al. (1968). Biotechnol. Bioeng., 10,845.
Aiba, S., et al. (1969a). AIChEJ, Amer. Inst. Chern. Eng. Journal, 15,624.
Aiba, S., et al. (1969b). Biotechnol. Bioeng., 11, 1285.
Andrews, J.F. (1968). Biotechnol. Bioeng., 10,707.
Andrews, J.F. (1969). J. Sanit. Eng. Div., ASCE, 95, 95.
Andrews, IF. (1971). Biotechnol. Bioeng. Symp., 2, 5.
Andreyeva, L.N., and Biryukov, V.V. (1973). Biotechnol. Bioeng. Symp., 4, 61.
Aris, R, and Humphrey, A.E. (1977). Biotechnol. Bioeng., 19, 1375.
Asai, T., and Kono, T. (1983). Adv. Ferment., 83 (Suppl. Proc. Biochem.), 212.
Asai, T., et al. (1978). J. Ferment. Technol., 56, 369.
Atkinson, B., et al. (1969). Trans. Inst. Chem. Engrs., 45, T257.
Aziz, J.A., and Tebbut, T.H.Y. (1980). Water Res., 14,319.
Bader, F.G. (1978). Biotechnol. Bioeng., 20,183.
Bader, F.G. (1982). In Bazin, M.J. (ed.). Microbial Population Dynamics. Boca Raton,
Fla.: CRC Press.
Bader, F.G., et al. (1975). Biotechnol. Bioeng., 17,279.
Bailey, J.E., and Ollis, D.F. (1977). Biochemical Engineering Fundamentals. New York:
McGraw-Hill.
Bajpaj, R.K. and Reuss M. (1980). J. Chem. Tech. Biotechnol., 30, 332.
Bajpaj, R.K., and Ghose, T.K. (1978). Biotechnol. Bioeng., 20. 927.
Bajpaj, R.K., and Reuss, M. (1980). Paper presented at 6th Internat. Fermentation
Symp., London, Ontario.
Bajpaj, R.K., and Reuss M. (1982). Biotechnol. Bioeng., 23, 717.
Barford, J.P., and Hall, R.J. (1979a). In Proceedings 7th Australian Conference on
Chemical Engineering, p. 27.
Barford, J.P., and Hall, RJ. (1979b). J. Gen. Microbiol., 114,267.
Barford, J.P., et al. (1982). In Bazin, M.J. (ed.). Microbial Population Dynamics. Boca
Raton, Fla.: CRC Press.
Barnes, R., Vogel, H., and Gordon, I. (1969). Proc. Nat. Acad. Sci. (Wash.), 62, 263.
Bayer, K., and Meister, G. (1982). Eur. J. Appl. Microbiol. Biotechnol., 15, 36.
Bazin, M.J. (1982). In Bushell, M.E., and Slater, J.H. (eds.). Mixed Culture Fermentations.
London~New York: Academic Press, Chap. 2.
Bazua, C.D., and Wilke, C.R (1977). Biotechnol. Bioeng. Symp., 5, 105.
Bergter, F. (1972). Wachstum von Mikroorganismen. Jena, East Germany; G. Fischer
Verlag.
Bergter, F. (1978). Z. Allgem. Mikrob., 18(2), 143.
Bergter, F., and Knorre, W. (1972). Z. Allgem. Mikrob., 12(8),613.
Bertalanffy, L., von (1942). Theoretische Biologie. Berlin: Borntraeger.
Blackman, F.F. (1905). Ann. Bot., 19,281.
Bleecken, St. (1979). Populationsdynamik einzelliger Mikroorganismen, Modellbildung
und -anwendung. (Fortschritte der Experimentellen und Theoretischen Biophysik),
vol. 23. (Beier W., ed.) Leipzig: Georg Thieme.
Bol, G., et al. (1980). In Proceedings of the 6th International Fermentation Symposium,
Adv. in Biotechn., vol. I, p. 339.
Borzani, W., and Baralle, S.B. (1983). Biotechnol. Bioeng., 25, 3201.
Borzani, W., and Hiss, H. (1979). Biotechnol. Bioeng., 21, 2149.
Botts, J., and Morales, M. (1953). Trans. Faraday Soc., 49, 696.
Boyle, W.e., and Berthouex, P.M. (1974). Biotechnol. Bioeng., 16, 1139.
Brettel, R., et al. (1981a). Eur. J. Appl. Microbiol. Biotechnol., 11,205.
Brettel, R., et al. (1981b). Eur. J. Appl. Microbiol. Biotechnol., 11,212.
Bronn, W.K. (1971). Chem. lng. Techn., 43, 70.
Brown, D.E., and Halsted, D.J. (1975). Biotechnol. Bioeng., 17, 1199.
Brown, D.E., and Vass, R.C. (1973). Biotechnol. Bioeng., 15,321.
Budd, J.A. (1977). Eur. J. Appl. Microbiol., 3, 267.
Bu'Lock, J.D. (1975). In Smith, J.E., and Berry, D.R. (eds.). The Filamentous Fungi,
Vol. 1. London: Edward Arnold Ltd., Chap. 3.
Bungay, H.R., III, and Bungay, M.L. (1968). Adv. Appl. Microbiol., 10,269.
Busby, J.B., and Andrews, J.F. (1975). J. Water Pollut. Contr. Fed., 47, 1055.
Calam, e.T., and Russell, D.W. (1973). J. Appl. Chem. Biotechnol., 23, 225.
Calam, e.T., et al. (1971). J. Appl. Chem. Biotechnol., 21, 181.
Caldwell, I.Y., and Trinci, A.P.J. (1973). Arch. Mikrobiol., 88,1.
Campbell, A. (1957). Bact. Rev., 21, 263.
Canale, R.P. (1969). Biotechnol. Bioeng., 11,887.
Canale, R.P. (1970). Biotechnol. Bioeng., 12, 353.
Chen, C.V. (1955). Chem. Rev., 55, 595.
Chen, Y.R., and Hasimoto, A.G. (1978). Biotechnol. Bioeng. Symp., 8, 269.
Chiu, Y.S., and Zajic, J.E. (1976). Biotechnol. Bioeng., 18, 1167.
Chiu, S.Y., et al. (1972). Biotechnol. Bioeng., 14, 179.
Constantinides, A., et al. (1970a). Biotechnol. Bioeng., 12, 803.
Constantinides, A., et al. (1970b). Biotechnol. Bioeng., 12, 1081.
Contois, D.E. (1959). J. Gen. Microbiol., 21,40.
Cooney, Ch.L., et al. (1969). Biotechnol. Bioeng., 11, 269.
Crooke, Ph.S., and Tanner, R.D. (1980). Appl. Math. Modelling, 4, 376.
Curds, e.R. (1971). Water Res., 5, 793.
Cysewski, G.R., and Wilke, Ch.R. (1978). Biotechnol. Bioeng., 20,1421.
Czerkawski, J.W. (1973). Process Biochem., October, 25.
Dabes, J.N., Finn, R.K., and Wilke, C.R. (1973). Biotechnol. Bioeng., 15, 1159.
Dagley, S., and Hinshelwood, e.N. (1938). J. Chem. Soc., 1930.
Dean, A.C.R., and Hinshelwood, C.N. (1966). Growth, Function and Regulation in
Bacterial Cells. Oxford: Clarendon Press.
Demain, A.L., et aI. (1979). In 29th Symposium of the Society of General Microbiology,
Cambridge: Cambridge University Press. (Bull A., ed.).
Dialer, K., and Lowe, A. (1975). In Winnacker, K., and Kuchler, L. (eds.). Chemische
Technologie, Vol. 7, 3rd ed., Munich: e. Hanser.
Eckenfelder, W.W., and Ford, D.L. (1970). Water Pollution Control. Austin, Tex.:
Pemberton Press.
Edwards, V.H. (1969). Biotechnol. Bioeng., 11, 99.
Edwards, V.H. (1970). Biotechnol. Bioeng., 12,679.
Edwards, H.V., and Wilke, Ch.R. (1968). Biotechnol. Bioeng., 10, 205.
Eggers, E. and Terlouw, T. (1979). Water Res., 13, 1077.
Esener, A.A., et al. (1982a). In Proceedings of the 2nd IWTU, Intern. Symp. Waste
Treatment Utilization. Moo-Young, M., et al. (eds.).: Pergamon Press, Oxford,
New York Vol. 2.
Esener, A.A., et al. (1982b). Biotechnol. Bioeng., 24,1749.

Falch, E. (1968). Biotechnol. Bioeng., 10,233.
Fennema, Q.F. (1975). Principles of Food Science, Part II. New York and Basel: Marcel
Dekker, p. 54ff.
Fidgett, M., and Smith, E.L. (1975). J. Appl. Chem. Biotechnol., 25, 355.
Fiechter, A. (1982). In Biotechnology-A Comprehensive Treatise (Rehm, H.J., and
Reed, G., eds.) Verlag Chemie, Weinheim, 1, Chap. 7.
Fishman, V.M., and Biryukov, V.V. (1973). Biotechnol. Bioeng. Symp., 4, 647.
Follman, H., Markl, H., and Vortmeyer, D. (1977). Dechema Annual Conference,
June 23-24, Frankfurt, Lecture No. 9.15.
Forrest, W.W. (1972). In Norris, lR, and Ribbons, D.W. (eds.). Methods in Microbiology,
Vol. 6.B. London-New York: Academic Press, p. 285.
Frederickson, A.G., et al. (1970). Adv. Appl. Microbiol., 13,419.
Frouws, M.J., Vellenga, K., and Dewilt, H.G.J. (1976). Biotechnol. Bioeng., 18,53.
Fujimoto, Y. (1963). J. Theoret. Bioi., 5, 171.
Fukuda, H., et al. (1978). J. Ferment. Technol., 56, 351.
Gaden, E.L., Jr. (1955). Chem. Ind. (Lond.), 154, 192.
Gaden, E.L., Jr. (1959). J. Biochem. Microbiol. Technol. Eng., 1,413.
Garfinkel, D. (1969). In Concepts and Models in Biomathematics (Heinmets F., ed.)
New York: Marcel Dekker p. 1.
Gaudy, A.F., and Gaudy, E.T. (1972). Adv. Biochem. Eng., 2, 97.
Gause, G.F. (1934). The Struggle for Existence. Baltimore: Williams & Wilkins.
Ghose, T.K., and Tyagi, RD. (1979). Biotechnol. Bioeng., 21, 1401.
Giona, A.R., et al. (1976). Biotechnol. Bioeng., 18,473.
Giende, M., and Reich, J.G. (1972). Acta Bioi. Med. Germ., 29, 595.
Gottschal, J.C, and Thingstad, T.F. (1982). Biotechnol. Bioeng., 24, 1403.
Grady, C.P.L., et al. (1972). Biotechnol. Bioeng., 14,391.
Grant, CA., et al. (1978). Biotechnol. Bioeng., 19, 1817.
Grau, P., Dohanyos, M., and Chudoba, J. (1975). Water Res., 9, 637.
Grm, B., et al. (1980). Biotechnol. Bioeng., 22, 255.
Gutke, R. (1980). In UNEPjUNESCO/ICRO training course, Theoretical Basis of
Kinetics of Growth, Metabolism and Product Formation of Microorganisms. Jena:
Science Academy of East Germany, Central Institute for Microbiology and Experi-
mental Therapy (ZIMET), Vol. 1, pps. 30, and 112.
Gutke, R (1982). In Biotechnological Fundamentals of Continuous Biomass Pro-
duction, Manual for International Training Course. Jena (ZIMET) and Leipzig
(Inst. Techn. Chern.) East Germany: Academy of Science, p. 39 and p. 58.
Gutke, R, et al. (1980). Biotechnol. Lett., 2, 315.
Hamilton, B.K., Gardner, C.R., and Colton, C.K. (1974). AIChEJ, 20, 503.
Harder, W., and Dijkhuizen, L. (1975). In Dean, A.CR., et al. (eds.). Continuous Culture,
Vol. 6. London: SCI; Chichester: E. Horwood, Chap. 23, p. 297.
Harder, A., and RoeIs, J.A. (1982). Adv. Biochem. Eng., 21, p. 56.
Harremoes, P. (1977). Vatten,2, 122.
Harremoes, P. (1978). In Mitchell, R. (ed.) Water Pollution Microbiology, Vol. 2. J.
Wiley, N.Y. p. 71.
Harris, R.S., and Karmas, E. (1975). Nutritional Evaluation of Food Processing, 2nd ed.
Westport, Conn.: AVI Publishing, p. 211ff.
Han, K. and Levenspiel, O. (1987). Biotechnol. Bioeng. 30, in press.

Hartmeier, W., Bronn, W.K., and Dellweg, H. (1971). Chern. lng. Techn., 43, 76.
Heckershoff, H., and Wiesmann, U. (1981). Chern. lng. Techn., 53, 268.
Hegewald, E.M., et al. (1978). In Sikyta, B. et aI., eds. Proceedings of the 7th Symposium
on Continuous Culture of Microorganisms. Czech. Acad. Sciences Prague: p. 717
(1980).
Hegewald, E.M., et al. (1981). Biotechnol. Bioeng., 23,1563.
Heidel, D., and Schultz, J.S. (1978). Biotechnol. Bioeng., 20, 30l.
Heijnen, J.J., et al. (1979). Biotechnol. Bioeng., 21, 2175.
Heineken, F.G., and O'Connor, R.J. (1972). J. Gen. Microbiol., 73, 35.
Heinzle, E., et al. (1982). In Proc. IFAC Symposium "Modelling & Control of Bio-
technical processes" (Halme, A., ed.) p. 57 Pergamon Press, Oxford.
Herbert, D. (1958). In Malek, I., ed., Proceedings of the 1st Symposium on Continuous
Culture of Microorganisms. Publ. House Czech. Acad. Sciences Prague: p. 45.
Herbert, D. (1961). In Elsworth, R. ed. Proceedings of the 2nd Symposium on "Con-
tinuous Culture of Microorganisms." Soc. Chern. Ind. London: SCI Monograph, 12,
p.2l.
Hess, B. (1973). In Symposium of the Society of Experimental Biology. Cambridge:
Cambridge University Press, p. 105.
Hess, B., and Boiteux, A. (1973). In Change, B., Pye, E.K., Ghosh, A.K., and Hess, B.
(eds.). Biological and Biochemical Oscillations. New York: Academic Press.
Hill, A.V. (1910). J. Physiol. (Lond.), 40, 4.
Higgins, J. (1967). Ind. Eng. Chern., 59, 19.
Hinger, K.J., and Blenke, H. (1975). Chern. lng. Techn., 47, 976.
Hinshelwood, CN. (1946). The Chemical Kinetics of the Bacterial Cell. Oxford:
Clarendon Press.
Hiss, H., and Borzani, W. (1983). Biotechnol. Bioeng., 25, 3079.
Holzberg, I., et al. (1967). Biotechnol. Bioeng., 9, 413.
Hoppe, G.K., and Hansford, G.S. (1982). Biotechnol. Lett., 4, 39.
Hornby, W.E., Lilly, M.D., and Crook, E.M. (1968). Biochem. J., 107,669.
Howell, J.A., and Atkinson, B. (1976). Biotechnol. Bioeng., 18, 15.
Humphrey, A.E. (1972). Chern. Reaction Eng., Adv. Chern. Ser., 109,603.
Humphrey, A.E. (1977a). Encyc/. Chern. Proc. Des., 4, 359.
Humphrey, A.E. (1977b). Chern. Eng. Prog., 85.
Humphrey, A.E. (1978). Am. Chern. Soc. Symp. Ser., 72.
Humphrey, A.E., and Yang, R.D. (1975). Biotechnol. Bioeng., 17, 121l.
Imai, H., et al. (1979a). J. Ferment. Technol., 57, 333.
Imai, H., et al. (1979b). J. Ferment. Technol., 57, 453.
Imanaka, T., and Aiba, S. (1976). J. Appl. Chern. Biotechnol., 26, 559.
Imanaka, T., and Aiba, S. (1977). Biotechnol. Bioeng., 19,757.
Imanaka, T., et al. (1972). J. Ferment. Technol., 50, 633.
Imanaka, T., Kaieda, T., and Taguchi, H. (1973) J. Ferment. Techno!., 51, 423.
Jannasch, H.W., and Mateles, R.I. (1974). Adv. Microbiol. Physiol., 11, 165.
Jerusalimsky, N.D. (1967). In Powell, E.O., et al. (eds.). Microbial Physiology and
Continuous Culture. London: Her Majesty's Stationery Office, p. 23.
Jerusalimsky, N.D., and Neronova, N.M. (1965). Dokl. AN USSR, 161, 1437 (Russ.).
Johnson, F.H., Eyring, H., and Polissar, M.J. (1954). The Kinetic Basis of Molecular
Biology. New York: John Wiley.
Jones, R.P., and Greenfield, P.F. (1981). Biotechnol. Lett., 3, 225.

Joschek, H.I., Dehler, J., Koch, W., Engelhardt, H., and Geiger, W. (1975). Chem. Ing.
Techn., 47, 422.
Jost, J.L., et al. (1973). J. Theoret. BioI., 41, 46l.
Kafarow, W.W. (1971). Kybernetische Methoden in der Chemie und Chemischen Tech-
nologie. Berlin: Akademie Verlag.
Kafarow, W.W., Winarow, A.J., and Gardejew, L.S. (1979). Modelling of Biochemical
Reactors. Moscow: Lesnaja Promyshlenost (Russ.).
Kargi, F. (1977). J. Appl. Chem. Biotechnol., 27, 704.
Kargi, F., and Shuler, M.L. (1979). Biotechnol. Bioeng., 21, 187l.
Kawashima, E. (1973). J. Ferment. Technol., 51, 4l.
Kendall, D.G. (1949). J. Roy. Statist. Soc., B11, 230.
Kessick, M.A. (1974). Biotechnol. Bioeng., 16, 1545.
Kharash, M.S. (1929). Bur. Stand. J. Res., 2, 359.
Knorre, W.A. (1968). In Malek, I., Beran, K., Fencl, Z., Munk, V., Ricica, J., and
Smrckova, H. (eds.) Gontinuous Cultivation of Microorganisms, Proceedings of 4th
Symposium, Prague, New York: Academic Press, p. 225.
Knorre, W.A. (1976). In Mathematische Modellbildung in NaturwissenschaJt und Technik.
Berlin: Akademie Verlag, p. 22l.
Knorre, W.A. (1980). In Beier, W., and Rosen, R. (eds.). Biophysikalische Grundlagen
der Medizin. Stuttgart, New York: G. Fischer Verlag, p. 132.
Knorre, W.A., et al. (1978a). Zeit. Allgem. Mikrob., 18(4), 255.
Knorre, W.A., et al. (1978b). Zeit. Allgem. Mikrob., 18(8),609.
Knowles, G., et al. (1965). J. Gen. Microbiol., 38, 263.
Kobayashi, T., and Moo-Young, M. (1971). J. Gen. Microbiol., 13, 893.
Kobayashi, T., et al. (1973). Biotechnol. Bioeng., 15,27.
Koga, S., et al. (1967). Appl. Microbiol., 15,683.
Konak, A.R. (1974). J. Appl. Chem. Biotechnol., 24, 453.
Kono, T. (1968). Biotechnol. Bioeng., 10, 105.
Kono, T., and Asai, T. (1968a). J. Ferment. Technol., 46, 391.
Kono, T., and Asai, T. (1968b). J. Ferment. Technol., 46, 398.
Kono, T., and Asai, T. (1968c). J. Ferment. Technol., 46, 406.
Kono, T., and Asai, T. (1969a). Biotechnol. Bioeng., 11, 19.
Kono, T., and Asai, T. (1969b). Biotechnol. Bioeng., 11,293.
Kono, T., and Asai, T. (1969c). J. Ferment. Technol., 47,651.
Kono, T., and Asai, T. (1971a). J. Ferment. Technol., 49, 128.
Kono, T., and Asai, T. (1971b). J. Ferment. Technol.,49, 133.
Kosaric, N., et al. (1984). Acta Biotechnolog., 4, 153.
Kossen, N.W.F. (1979). Symp. Soc. Gen. Microbiol., 20, p. 327.
Kossen, N.W.F., and Roe1s, J.A. (1978). In Bull, M.J., ed., Progress of Industrial
Microbiology, Vol. 14. Elserrer 95.
Kremen, A. (1971). J. Theoret. Bioi., 31, 363.
Krotzsch, P., Kurten, H., Daucher, H., and Popp, K.H. (1976). Chem. Ing. Techn.,
MS,351.
Labuza, Th.P. (1980). Food Technol., February, 67.
La Motta, E.J. (1976). Biotechnol. Bioeng., 18, 1029, 1359.
Lee, I.H. (1973). Ph.D. Thesis, University of Minnesota, Minneapolis.
Lee, I.H., et al. (1976). Biotechnol. Bioeng., 18,513.
Lerner, J.E. (1966). J. Org. Chern., 31, 533.

Levenspiel, O. (1972) Chemical Reaction Engineering, 2nd ed., John Wiley & Sons,
New York.
Levenspiel, O. (1980). Biotechnol. Bioeng., 22, 1671.
Lodge, R.M., and Hinshelwood, C.N. (1943). J. Chern. Soc., 213.
Lotka, A.J. (1925). Elements of Mathematical Biology. New York: Dover. (Reprinted
1956.)
Lovrien, R., et al. (1980). Biotechnol. Bioeng., 22, 1249.
Luedeking, A., and Piret, E.L. (1959). Biotechnol. Bioeng., 1, 393.
Lumry, R., and Eyring, H. (1954). J. Phys. Chern., 58, 110.
Luong, J.H.T. (1982). Eur. J. Appl. Microb. Biotechnol., 16,28.
Luong, J.H.T., and Volesky, B. (1980). Canad. J. Chern. Eng., 60,163.
Maaloee, 0., and Kjeldgaard, N.O. (1965). The control of Macromolecular Biosynthesis.
New York: Benjamin.
Mahler, H.R., and Cordes, E.H. (1966). Biological Chemistry. New York: Harper
International.
Marr, A.G. et al. (1963). Ann. N. Y. Acad. Sci., 102, 536.
Mason, T.J., and Millis, N.F. (1976). Biotechnol. Bioeng., 18, 1337.
May, R.M. (1973). Stability and Complexity in Model Ecosystems. Princeton, N.J.:
Princeton University Press, p. 25.
Meers, J.L. (1973). CRC Crit. Rev. Microbiol., 2,139.
Megee, R.D., Drake, J.F., Fredrickson, A.G., and Tsuchiya, H.M. (1972). Canad. J.
M icrobiol., 18, 1733.
Megee, R.D., III, et al. (1970). Biotechnol. Bioeng., 12, 771.
Metz, B. (1975). Ph.D. Thesis, Technical University, Delft, Netherlands.
Metz, B., and Kossen, N.W.F. (1977). Biotechnol. Bioeng., 19,781.
Meyer, J.S., et al. (1975). Biotechnol. Bioeng., 17, 1065.
Meyrath, J. (1973). Mitteil. Versuchsanstalt fur Giirungsgewerbe Wien, 5/6,95.
Meyrath, J., and Bayer, K. (1973). In Dellweg, H. (ed.). Proceedings of the 3rd Symposium
Technical Microbiology. Berlin: Inst. f. Giirungsgewerbe und Biotechnologie, p. 117.
Minkevich, I.G., and Eroshin, V.K. (1973). Folia Microbiol., 18,376.
Miura, Y., et al. (1980). Biotechnol. Bioeng., 22, 929.
Mochan, E., and Pye, E.K. (1973). Nature New Bioi., 242,177.
Mona, R. (1978). Ph.D. Thesis, Technical University, Zurich, No. 6088.
Mona, R., et al. (1979). Biotechnol. Bioeng., 21,1561.
Monk, P.R. (1978). Proc. Biochem., December, 4.
Hermann & Cie.
Monod, J. (1949). Ann. Rev. Microb., 3, 371.
Monod, J., Changeux, J.P., and Jacob, F. (1963). J. Molec. Bioi., 6, 306.
Moreira, A.R., et al. (1979). Biotechnol. Bioeng. Symp., 9, 179.
Morgan, M.S., and Edwards, V.H. (1971). Chern. Eng. Progr. Symp., Ser., 114,67,51.
Moser, A. (1974). Gas Wasserfach/Wasser-Abwasser, 115,411.
Moser, A. (1977). Thesis of habilitation Technical University, Graz, Austria.
Moser, A., Lafferty, R.M. (1977). Zbl. Bakt. I. Abt. Ref., 252, 60.
Moser, A. (1978a). In Proceedings of the 1st European Congress on Biotechnology,
Interlaken, Switzerland, Part 1, p. 88.
Moser, A. (1978b). Gas Wasserfach/Wasser-Abwasser, 119,242.
Moser, A. (1978c). In P. Galzy et aI., Chair of Genetics & Microbiology (ed.) Proceedings
of the 6th International Special Symposium on Yeasts, July 2-8 Montpellier, France,
p. 16. Administrative Secretariat Midi Contacts.
Moser, A. (1980a). UNEPjUNESCOjICRO training course, Theoretical Basis of
Kinetics of Growth, Metabolism and Product Formation of Microorganisms. Jena:
Science Academy of East Germany, Central Institute for Microbiology and Experi-
mental Therapy (ZIMET), Part II, p. 27.
Moser, A. (1980b) In Moo-Young, M., et al. (eds.). Proceedings of the 2nd International
Symposium on Waste Treatment and Utilization. Oxford, New York: Pergamon
Press, Vol. 2, 177.
Moser, A. (1981). Bioprozesstechnik. Vienna-New York: Springer-Verlag.
Moser, A. (1983a). Acta Biotechnolog., 3, 195.
Moser, A. (1983b). In Proceedings Biotechnology '83, International Conference on Com-
mercial Applications and Implications of Biotechnology. London: Online, Northwood
p.961.
Moser, A. (1984a). In Rehm, H.J., and Reed, G. (eds.). Biotechnology-A Comprehensive
Treatise, Vol. 2. Deerfield Beach, Fla., and Basel, Verlag Chemie Weinheim, Chap.
16.5.
Moser, A., and Schneider, H. (1988) Bioprocess Engineering 4, in press.
Moser, A. (1984b). Acta Biotechnolog. 4,3.
Moser, A., Kosaric, N., and Margaritis, A. (1980b). Paper at 30th Canad. Chern. Eng.
Conf., Edmonton, Alberta, 19-22 October.
Moser, A., and Steiner, W. (1975). Eur. J. App/. Microbiol., 1,281.
Moser, A. (1980). Paper at 6th Internat. Ferm. Symp., London, Ontario. July 20-25.
Moser, F. (1977). Verfahrenstechnik, 11,670.
Moser, H. (1958). The Dynamics of Bacterial Populations Maintained in the Chemostat.
Carnegie Institution, Washington D.C. Publ. no. 614.
Moo-Young, M., and Kobayashi, T. (1972). Canad. J. Chem. Eng., 50, 162.
Mou, D.G., and Cooney, Ch.L. (1976). Biotechnol. Bioeng., 18, 1371.
Netter, H. (1969). Theoretical Biochemistry. Edinburgh: Oliver & Boyd.
Nino, G.J., and Ross, L.W. (1969). Biotechnol. Bioeng., 11,719.
Noack, D. (1968). Biophysikalische Prinzipien der Populationsdynamik in der Mikro-
biologie (F ortschritte der experimentellen und theoretischen Biophysik Vol. 8, Beier,
W., ed.) Leipzig: Georg Thieme.
Nyholm, N. (1976). Biotechnol. Bioeng., 18, 1043.
Ohtaguchi, K., et al. (1979). Internat. Chem. Eng., 19,313,591.
Oleskiewicz, J.A. (1977). Prog. Water Technol., 9, 777.
Oura, E. (1973). In Patomaki, L., and Kiuru, A. (eds.). Proceedings of the 1st National
Meeting. Biophysics and Biotechnology. Finland: ALKO, Helsinki p. 142.
Pamment, N.B., et al. (1978). Biotechnol. Bioeng., 20, 349.
Pardee, A.B. (1961). Symp. Soc. Gen. Microb., 11, 19.
Peringer, P., et al. (1974). Biotechnol. Bioeng., 16,431.
Perret, c.J. (1960). J. Gen. Microbiol., 22, 589.
Peters, H. (1976). In Hartmann, L. (ed.). Karlsruher Berichte zur Ingenieurbiologie.
Karlsruhe: Inst. f. Ingenieurbiologie und Biotechnologie, p. 9. Univ. Karlsruhe.
Petrova, T.A., et al. (1977). Z. Allgem. Mikrob., 17(7),531.
Pickett, A.M. (1982). In Bazin, M.J. (ed.). Microbial Population Dynamics. Boca Raton,
Fla.: CRC Press, Chap. 4.
Pirt, S.J. (1965). Proc. Roy. Soc. B., 163,224.

Pirt, S.J. (1975). Principles of Microbe and Cell Cultivation. Oxford: Blackwell Scientific
Publishing.
Pirt, S.J., and Righelato, R.C (1967). Appl. Microbiol., 15, 1284.
Pirt, S.J., et al. (1983). J. Chem. Tech. Biotechnol., 33B, 35.
Pitcher, W.H. (1978). Adv. Biochem. Eng., 10, 1.
Powell, E.O. (1958). J. Gen. Microbiol., 18,259.
Powell, E.O. (1967). In Powell, E.O., et al. (eds.). Microbial Physiology and Continuous
Culture. London: Her Majesty's Stationery Office, p. 34.
Prokop, A., and Humphrey, A.E. (1970). In Bernardo, M.A. (ed.). Disinfection. Marcel
Dekker, New York: p. 61.
Prosser, 1.1. (1982). In Bazin, M.J. (ed.). Microbial Population Dynamics. Boca Raton,
Fla.; CRC Press, Chap. 5.
Ramkrishna, D., et al. (1966). J. Ferment. Technol.,44, 210.
Ramkrishna, D., et al. (1967). Biotechnol. Bioeng., 9, 129.
Ratnam, D.A., et al. (1982). Biotechnol. Bioeng., 24, 2675.
Reich, J.O., et al. (1972). Eur. J. Biochem., 26, 368.
Reilly, P.J. (1974). Eur. J. Biochem., 16, 1373.
Reuss, M., and Wagner, F. (1973). Proceedings of the 3rd Symposium on Technology
and Microbiology. Berlin Dellweg, H., ed., p. 89, Inst. f. Garungsgewerbe und
Biotechnologie.
Reuss, M. (1977). Fort. Verfahrenstechnik, 15F, 549.
Reuss, M., and Bajpaj, R.K. (1982) J. Chem. Technol. Biotechnol. 32, 81.
Ricica, J. (1969). In Perlman, D. (ed.). Fermentation Advances. New York: Academic
Press, p. 427.
Rittmann, B.E., and McCarty, P.L. (1981). J. Environ. Eng. Div., ASCE, 107(EE4), 831.
Roels, JA (1978). Proceedings of the 1st European Congress on Biotechnology, Interlaken,
Switzerland, Dechema Monograph, 82, 221.
Roels, JA (1980). Biotechnol. Bioeng., 22, 2457.
Roels, JA (1982). J. Chem. Tech. Biotechnol., 32, 59.
Roels, J.A. (1983). Energetics and Kinetics in Biotechnology. Amsterdam, New York,
Oxford: Elsevier Biomedical Press.
Roels, J.A., and Kossen, N.W.F. (1978). In Bull, M.J. (ed.). Progress in Industrial
Microbiology, Vol. 14. Amsterdam: Elsevier, p. 95.
Romanovsky, J.M., et al. (1974). Kinetische Modelle in der Biophysik. Jena: G. Fischer.
(German translation by W.A. Knorre and A. Knorre).
Rowley, B.I., and Pirt, S.J. (1972). J. Gen. Microbiol., 72, 553.
Ryder, D.N., and Sinclair, CG. (1972). Biotechnol. Bioeng., 14,787.
Ryu, D.Y., and Humphrey, A.E. (1972). J. Ferment. Technol., 50,424.
Ryu, D.Y., and Mateles, R.I. (1968). Biotechnol. Bioeng., 10,385.
Saguy, I., and Karel, M. (1980). Food Technol., February, 78.
Schaezler, D.J., McHarg, W.H., and Busch, A.W. (1971). Biotechnol. Bioeng. Symp.,
2,107.
Schmid, R., and Sapunov, V.N. (1982). Non-Formal Kinetics. Monographs in Modern
Chemistry, Vol. 14. Deerfield Beach, Fla., and Basel: Verlag Chemie Weinheim.
Schneider, H., Moser, A. (1986). Bioprocess Engng. 2, 129.
Schiigerl, K. (1983). In Proceedings, NATO ASI, Mass Tranifer with Chemical Reactions,
Izmir Turkey, 1981, Vol. 1(72), p. 415.
Shuler, M.L., et al. (1972). J. Theoret. Bioi., 35, 67.

Schultz, J.s., et aI. (1974). AIChEJ, 20,417.
Schultz, K.L., et al. (1964). Arch. Mikrobiol., 48, l.
Segel, I.H. (1975). Enzyme Kinetics. New York: Wiley-Interscience.
Setzermann, u., et al. (1982). Acta Biotechnolog., 2, 325.
Shehata, T.E., and Marr, A.G. (1971). J. Bacteriol., 107, 210.
Sheintuch, M. (1980). Biotechnol. Bioeng., 22, 2557.
Shieh, W.K. (1980a). Biotechnol. Bioeng., 22, 667.
Shieh, W.K. (1980b). Water Res., 14,695.
Shu, P. (1961). Biotechnol. Bioeng., 3, 95.
Shuler, M.L., and Domach, M.M. (1982). In Blanch, H.W., et al. (eds.). Amer. Chern.
Soc. Winter Symp. Kinetics and Thermodynamics in Biological Systems. Boulder,
Colo.: Amer. Chern. Soc.
Shuler, M.L., et al. (1979). Ann. N. Y. Acad. Sci., 326, 35.
Siimer, E. (1978). Biotechnol. Bioeng., 20,1853.
Sikyta, B., et aI. (1977). In Meyrath, J., and Bu'Lock, J.D. (eds.). FEMS Symposium 4
New York: Academic Press, 119.
Sinclair, e.G., and Topiwala, H.H. (1970). Biotechnol. Bioeng. 12, 1069.
Sonnleitner, B., and Fiechter, A. (1983). Trends in Biotechnol. 1, 74.
Stephanopoulos, G., and Fredrickson, A.G. (1979). Biotechnol. Bioeng., 21, 149l.
Stephanopoulos, G., et al. (1979). AIChEJ, 25,863.
Sudo, R., et aI. (1975). Biotechnol. Bioeng., 17, 167.
Suga, K., et al. (1975). Biotechnol. Bioeng., 17, 185.
Sykes, R.M. (1973). J. Water Pollut. Contr. Fed., 45, 888.
Taguchi, H., et aI. (1978). J. Ferm. Technol., 56,158.
Takamatsu, S., et aI. (1974). J. Ferment. Technol., 52,190.
Takamatsu, T., et al. (1981). J. Ferment. Technol., 59, 13l.
Talsky, G. (1971). Angew. Chem., 83, 553.
Tanaka, H., Uzman, S., and Dunn., I.J. (1981). Biotechnol. Bioeng., 23,1683.
Tanner, R.D. (1970). Biotechnol. Bioeng., 12, 83l.
Teissier, G. (1936). Ann. Physiol. Physiochim. Bioi., 12, 527.
Tempest, D.W. (1976). In Continuous Culture 6: Applications and New Fields (Dean,
A.C.R. et aI., eds.) Society of Chemical Industry, London: Chichester: E. Harwood
Ltd., p. 349.
ten Hoopen, H.J.G., et aI. (1980). Adv. Biotechnol., 1,315.
Terui, G. (1972). In Sterbacek, Z. (ed.). Microbial Engineering. London: Butterworth,
p.377.
Theophilou, J., Wolfbauer, 0., and Moser, F. (1978). Gas-WasserFachjWasser-Abwasser
119,135.
Thijssen, H.A.e. (1979). Lebensm.-Wiss. -Technol., 12, 308.
Toda, K. (1976). Biotechnol. Bioeng., 18, 1117.
Toda, K. (1980). Biotechnol. Bioeng., 22,1805.
Toda, K. (1981). J. Chem. Tech. Biotechnol., 31, 775.
Topiwala, H.H., and Hamer, G. (1971). Biotechnol. Bioeng., 13,919.
Trilli, A. (1977). J. Appl. Chem. Biotechnol., 27, 25l.
Trinci, A.P.J. (1970) Arch. Microbiol., 73, 353.
Tsao, G.T., and Hanson, Th.P. (1975). Biotechnol. Bioeng., 17, 159l.
Tsao, G.T., and Yang, C.M. (1976a). Biotechnol. Bioeng., 18, 1827.
Tsao, G.T., and Yang, C.M. (1976b). In Dellweg, H., ed. Proc. 5th Intemat. Fenn.
Symp., Berlin, Abstract No. 5.03. Inst. f. Giirungsgewerbe und Biotechnologie,
June 28-July 3.
Tseng, M., and Phillipps, C.R. (1981). Biotechnol. Bioeng., 23,1639.
Tseng, M.e., and Wayman, M. (1975). Canad. J. Microbiol., 21, 994.
Tsuchiya, H.M., et al. (1966). Adv. Chem. Eng., 6, 125.
Tsuchiya, H.M., et al. (1972). J. Bacteriol., 110, 1147.
Tucek, F., Chudoba, J., and Madera, V. (1971). Water Res., 5, 647.
van Dedem, G. (1975). Biotechnol. Bioeng., 17, 1301.
van Dedem, G., and Moo-Young, M. (1973). Biotechnol. Bioeng., 15,419.
van Suijdam, J.C., and Metz, B. (1981). Biotechnol. Bioeng., 23,111.
van Suijdam, J.C., et al. (1982). Biotechnol. Bioeng., 24, 177.
van Uden, N. (1971). Z. Allgem. Mikrob., 11(6),541.
van Uden, N., Abranches, P., and Cabeca-Silva, C. (1968). Arch. Microbiol., 61, 381.
van Uden, N., and Vidal-Leiria, M.M. (1976). Arch. Microbiol., 108,293.
Vavilin, V.A. (1982). Biotechnol. Bioeng., 24, 1721,2609.
Volesky, B., et al. (1982). J. Chem. Tech. Biotechnol., 32, 650.
Veldkamp, H. (1975). In Dean, A.C.R. (ed.). Continuous Culture, Vol. 6. London: SCI,
Chap. 24.
Veldkamp, H., and Jannasch, H.W. (1972). J. Appl. Chem. Biotechnol., 22,105.
Verhoff, F.H., et al. (1972). Biotechnol. Bioeng., 14,411.
Verhulst P.F. (1845). Recherches Mathematiques sur la Loi d'Accroissement de la
Population. Nouv. Mem. de l'Acad. Roy. des Sciences et Belles-Lettres de Bruxelles,
18, p. 1.
Volterra, V. (1931). Lecons sur la Theorie Mathematique de la Lutte pour la Vie. Paris:
Gauthier-Vallars.
von Bertalanffy, L. (1932,1942). Theoretische Biologie, West Berlin: Bomtraeger.
von Bertalanffy, L., et al. (1977). Biophysik des Fliessgleichgewichtes. East Berlin:
Akademie Verlag.
Wayman, M., and Tseng, M.e. (1976). Biotechnol. Bioeng., 18, 383.
Webb, J.L. (1963). Enzyme and Metabolic Inhibitors, Vol. 1. New York: Academic
Press.
Webster, I.A. (1983). Biotechnol. Bioeng., 25, 2981.
Weissmann, J.C., and Benemann, J.R. (1979). Biotechnol. Bioeng., 21, 627.
Whitaker, A., and Long, P.A. (1973). Proc. Biochem., November, 27.
Wilder, C.T., et al. (1980). Proc. Biochem., 22, 89.
Wilderer, P. (1976). In Karls. Ber. Ingenieurbiol., Vol. 8, Hartmann, L., ed. Inst. f.
Ingenieurbiologie und Biotechnologie, Univ. Karlsruhe 1-145.
Wilderer, P., Engelmann, G., and Schmenger, H. (1977). Gas-Wasser-Fach/Wasser-
Abwasser, 118, 357.
Wilderer, P., and Hartmann, L. (1978). In Moderne Abwasserreinigungsverfahren
(Munch. Beitr. Abw., Fisch. Flussbiol.) Oldenburg Verlag Munchen Vol. 29, p. 9.
Williams, F.M. (1967). J. Theoret. Bioi., 15,90.
Williams, F.M. (1975). In Patten, B.V. (ed.). System Analysis and Simulation in Ecology,
Vol. 7. New York: Academic Press, Chap. 3, p. 197.
Wolfbauer, 0., Klettner, H., and Moser, F. (1978). Chem. Eng. Sci., 33, 953.
Wohrer W., and Rohr, M. (1981). Biotechnol. Bioeng., 23, 567.
Wuhnnann, K., Beust, F. von, and Ghose, T.K. (1958). Schweiz. Hydrol., 20, 284.
Yagil, G., and Yagil, E. (1971). Biophys. J., 11, 1l.

Yano, T., et al. (1966). Agr. Bioi. Chern. (Jap.), 30, 42.
Yoon, H., and Blanch, H.W. (1977). J. Appl. Chern. Biotechnol. Bioeng., 19, 1193.
Yo on, H., Klinzing, G., and Blanch, H.W. (1977). Biotechnol. Bioeng., 19, 1193.
Young, T.B., Bruley, D.F., and Bungay, A.R. (1970). Biotechnol. Bioeng., 12,747.
Young, T.B., and Bungay, A.R. (1973). Biotechnol. Bioeng., 15,377.
CHAPTER 6
Bioreactor Performance:
Process Design Methods
Bioprocess design, as suggested in Fig. 2.14 (including especially calculation

of the conversion for prediction of the mode of reactor operation that will
lead to optimal production), represents the stage where kinetic and bioreactor
data are integrated. In practical situations, the dominant problems are often
maintenance of sterility, improvement in the strain of microorganism, and
isolation of the product. All of these greatly affect economic considerations.
Once in operation, a plant using a stirred vessel can usually be modified only
with respect to the operating conditions: Changing to a different type of
reactor is not an option. In the planning of a new operation, selection of both
the optimal mode of operation and the optimal reactor type is a foremost
consideration. Selection will be even more important in the future, when large
volume/low priced processes become economically competitive (cf. Fig. 1.2).
The problems posed by the interaction of the reactor type and the physi-
ology of the organism have only slowly been recognized (Finn and Fiechter,
1979; Melling, 1977; A. Moser, 1983a and b), cf. Sect. 3.3.11 and Leegwater
et al. (1982).
The present chapter will deal with general methods that are suitable for
making predictions in developing a process on the basis of kinetic data usually
obtained from measurements made in a discontinuously operated stirred
vessel. The design equation may be written in a qualitative way as
(i = f(ri , HvTR,OTR,RTD,J resp. te) (6.1)
Equation 6.1 is a generalized form of the conservation of mass equation (cf.
Equ.2.3).
The influence of the OTR on kinetics (and therefore on productivity) was
indicated in Equs. 5.169 and 5.170. Therefore, we will discuss here primarily
those process engineering factors involved in various different types of reactors
and operations. Last but not least, some unconventional reactors such as
membrane (dialysis) reactors and synchronous culture techniques will be
discussed.
308 6. Bioreactor Performance: Process Design Methods
6.1 The Ideal Single-Stage, Constant-Volume Continuous

Stirred Tank Reactor, CSTR (Pseudohomogeneous
L-Phase Reactor Model)
The continuous culture technique using a homogeneous single-stage stirred
tank reactor was introduced long ago in bioprocessing. The mathematical
foundation of continuous culture theory has been given in the basic papers of
Monod (1942, 1950) and of Novick and Szilard (1950). Many reviews and
several books have appeared on the theoretical and practical aspects of this
technique (Dean et aI., 1972; Fiechter, 1982; Herbert et aI., 1956; Malek and
Fencl, 1966; Malek and Ricica, 1969; A. Moser, 1985a; Powell et al., 1967;
Ricica, 1973; Tempest, 1970).
6.1.1 PERFORMANCE OF THE CSTR WITH SIMPLE KINETICS

In a completely mixed continuous flow reactor, the composition is thought
to be uniform throughout the reactor, and is the same as in the exist stream
(cf. Fig. 3.30). Applying the law of conservation of mass (cf. Equ. 2.3) to a CSTR
yields Equ. 3.90 for the steady state. The dilution rate D is reciprocal to the
mean residence time of a fluid (t).
The material balance in general form becomes
Accumulation = input - output reaction (6.2)
and for biomass X and substrate S
dx
dt = D'x in - D'xex + rx (6.3a)
ds
dt = D . Sin - D . Sex - rs (6.3 b)
At steady state with x = Xex = X R and s = Sex = SR'

rx = Dx (6.4)
rs = D(Sin - s) (6.5)
As a consequence of Equ. 6.4, the definition of growth rate Il in a CSTR at
steady state is given by Il = D (Equ. 3.91). This fact is most important for
practical application of the CSTR in microbial culture techniques. It means
that this reactor configuration functions as a differential reactor (cf. Sect. 4.4.1),
llnd it enables the direct measurement of a biological reaction rate (e.g., growth
rate) without mathematical manipulation of the measurements.
The practical consequence from this statement is that Il can be controlled
by the amount of nutrient fed to a constant-volume CSTR. This system
is called "realstat" or "chemostat," a name that refers to the constant chemical
6.1 The Ideal Single-Stage Stirred Tank Reactor 309
environment characteristic of the steady state. The nutrient medium is de-

signed so that one essential nutrient is the limiting substrate.
However, the simplest type of continuous culture system is a "turbidistat."
In this device, the cell concentration in the vessel is maintained at a constant
value by monitoring the optical density of the culture. The disadvantage of
this simple technique is the difficulty in adequately monitoring the cell concen-
tration. The basic behavior of both types of control apparatus for CSTR
operation becomes clear by analyzing CSTR reactor performance using the
balance equation previously discussed. Introducing the simple Monod equa-
tion (cf. Equ. 5.38) and using a yield coefficient Yx1s (cf. Equ. 2.14a) result in
the following set of equations, which represent the basic behavior of CSTRs:
s=K s (J1.maxD)
- D
(6.6)
(6.7)
A plot of these equations is given in Fig. 6.1, the dilution rate D being used as
the variable in Fig. 6.1a and the mean residence time T in Fig. 6.1 b. Figure
6.1b is preferred by process engineers, while Fig. 6.la is used predominantly
by microbiologists.
A modified turbidostat scheme has recently been proposed, as have other
control schemes such as pH auxostat, CTR, OTR, and OUR control (Agrawal
and Lim, 1984).
The critical washout point (Deri! or De) can be estimated from
So
Deri! = J1.max K
s
+ So (6.8)
and is somewhat smaller than J1.max'
x:s x,S
t
2-
i
X
a b
t cnt. -1"
FIGURE 6.1. Theoretical behavior of a continuous culture of microorganisms with

simple Monod kinetics in a stirred vessel (chemostat) with Xo = 0: Stationary concen-
tration of the cells (x) or substrate (5") as a function of the rate of dilution, D = F IV (a)
or mean residence time T = liD (b). In (b), the case for Xo > 0 is also shown by dotted
lines.
The productivity r CSTR for continuous processing in a single CSTR is

(6.9)
where cj is the steady-state concentration of component i (x, S, or p) in a
bioreactor at a given dilution rate D. Substituting Equ. 6.6 into Equ. 6.9 for
biomass production gives
rx = D YXIs[so - Ks (Dc ~ D) ] '(6.10)

The productivity has been plotted as a function of D or t in Fig. 6.1. The
dilution rate corresponding to optimum productivity Dopt can be calculated
for a single CSTR by setting the first derivative of Equ. 6.10 equal to zero.
Thus
D = Dc [1 _(KsKs+
opt
So
)1/2J (6.11 )
As expected, this maximum productivity does not coincide with the point of
maximum conversion of substrate. This fact was previously pointed out in
Fig. 2.12 with batch processes.
Finally, the ratio of the productivity in a CSTR to that in a DCSTR
becomes, with some simplifications (Dc ~ I1ma., Ks/so ~ 0, xmaxlxo 1),
rX,CSTR
(6.12)
rX,DCSTR
Using this equation, it is clear that continuous operation in a CSTR is superior

to batch processing, especially when organisms with maximum growth rates
of more than about 0.2 are used. Inserting the Michaelis~ Menten equation
(Equ. 2.54) for a poison-free enzymatic reaction into the performance equation
of a single CSTR (Equ. 3.90) gives
~ (so - s)(Km + s)
tCSTR = ~ (6.13)
rmax'S
An equation that does allow a direct evaluation of the kinetic parameters
rmax and Km can be obtained by rearranging Equ. 6.13 into a linear form:
S ~
s= -Km + rmax--~-t
So - S
(6.14)
A plot of this linear form is shown in Fig. 6.2 (Levenspiel, 1979). To evaluate
the kinetic parameters I1max and Ks from a set of CSTR data, Equ. 6.6 can be
rearranged using Equ. 6.2 to give a linear form
1 I1max ~ 1
~=--t-- (6.15)
S Ks Ks
A graphical representation is in Fig. 6.3.
5 115
1150 ----------:,- ,
:
~~/
",/
_ _ _ _L:_ _ _ _ _ _ _ _ _ _ _ _ _ _ t
1/rmax tcrit
FIGURE 6.2. Linear plot for enzyme kinetics according to Equ. 6.14, which allows
estimation of kinetic parameters rmax resp. rS,ma. and Km from CSTR operation (Leven-
spiel, 1979).
FIGURE 6.3. Linear plot for Monod kinetics according to Equ. 6.15 in the fonn of a
double reciprocal plot (cf. Fig. 4.24c) applied to a CSTR with D = l/T = 11-, which is
suitable for parameter estimation (I1-ma. and Ks).
A special method for determining P-rnax is used of the washout conditions in

a CSTR (D Dc). The balance equation of biomass in this case with S Ks
during washout experiments is
x = Xo' exp(p-rnax - D)t (6. 16a)
or when transformed
lnx = lnxo + (P-max - D)t (6.l6b)
Thus, J-lmax can easily be calculated by performing nonlinear or linear regres-

sion on the lnx versus t data (e.g., Esener et al., 1981b). The performance
equation shows that everything-washout, optimum cell concentration, max-
imum production rate-depends on Ks and So.
In the case ofaCSTR withxo #- 0, the balance equation is (Levenspiel, 1979)
- (x - xo)[Y(so + Ks) - (x - xo)]

I1-max t = Y . So . x - x (x - Xo ) (6.17)
Since Xo #- 0, no washout occurs and there is no restriction on P-max . t.
6.1.2 PERFORMANCE OF THE CSTR WITH COMPLEX KINETICS
Modifications ofbiokinetic models can be included in the CSTR performance

equation.
6.1.2.1 Lysis and Maintenance

Modifications have been introduced to deal with the discrepancy between
model behavior and chemostat experiments. Including the specific lysis rate
kd and the maintenance coefficient ms (cf. Equs. 5.80 and 5.76), the enlarged
model is
dx
dt = /l(s)x - kd X - Dx (6.18)
ds 1
- = - _ . /l(s)x - msx - D(so - s) (6.19)
dt Yx1s
which has the following nonwashout steady state:
( 2 x,
- 2-
s) = ( Yxis D { So - Ks D + kd },
D + kd + ms Yxis /lmax - D - kd
X
D + kd
Ks-----==---- ) (6.20)
/lmax - D - kd
Hence, lysis as well as maintenance may cause the curve of biomass concen-
tration against the dilution rate D to approach zero as the dilution rate tends
to zero (see Figs. 6.4 and 6.5). Furthermore, the lysis causes a decrease in the
washout point
So
Dc -- /l max + Ks - kd (6.21)
so
instead of Equ. 6.8 for Monod's model.

The specific lysis rate can be determined by the analysis of batch experi-
ments, but the maintenance coefficient cannot be so determined. The estima-
tion of this coefficient is possible by analysis of chemostat experiments. The
theoretical basis for this is given with Equ. 6.20. According to this relation,
the plot of the quotient (so - s)/x versus the reciprocal dilution rate liD gives
a str.aight line:
1 So - 2S 1 1 ( kd )
Y = 2X = Yxis + 15 Yxis + ms (6.22)
We call the quotient xl(so - s) the phenomenological yield coefficient Y; it

equals the "true" yield coefficient if the lysis and maintenance rates fall. If for
different dilution rates D the steady-state values x and s are measured with
known So and kd' one can determine "true" yield coefficient Yxis as well as the
maintenance coefficient ms by linear regression on the basis of Equ. 6.22.
A more structured model using a viability concept has been presented in
Fig. 5.24. The complexity of maintenance has been stressed in the literature
by introducing the definition of "coefficient of apparently non-finalized sub-
strate consumption" (Goma et aI., 1979) and by reporting the difficulties
involved in experimental verification of maintenance in the case of incorrect
FIGURE 6.4. Computer simula- 1~--~--~---~--~---~--~---~--~---~-~~==~

ii,s '1
tion of CSTR operation show-
ing the steady-state concentra- CELL LYSIS
tions (x and s) in dependence of
dilution rate D with variation
of specific rate of cell lysis kd
according to Equs. 6.18 and 6.19, kd
with ms = O. Parameter values: o 2-;
0.5 --.:oiOI;=:::=====+='=--.L
I1max = 0.8 h- 1 ; YXls = 0.5; Ks =
0.01 g'l-l; So = 1 gl-l.
';
0.5
--0
it,s
Maintenance metabolism
'.
.5
FIGURE 6.5. Effect of mainte-
nance metabolism on behavior
ofCSTR as seen by the influence
of maintenance coefficient ms
on steady-state concentrations
(x, s) as a function of dilution
'i/
rate D, with kd = O. (See Equs.
6.18 and 6.19.) -0
modeling by neglect of product formation during anaerobic growth (Esener

et al., 1981b). Other interesting papers concerning maintenance can be found
in literature (Kuhn et al., 1980; Solomon and Erickson, 1981).
6.1.2.2 Incomplete Mixing and Wall Growth

Another deviation often observed is the so-called wall growth, which results
in plots similar to Fig. 3.12. Topiwala and Hamer (1971) analyzed this effect
of the adherence of microorganisms to glass or metal surfaces. When the part
of adhered biomass that cannot wash out is denoted by x w , the following
modification of Monod's model results:
FIGURE 6.6. Effect of wall growth

x,s WALL GROWTH
on CSTR behavior in a plot of
.01
x and s versus D with varied
values of XW' (See Equ. 6.23.)
.5
l~======~~
o .5
__~==~
.... D
dx
It = I1(S)X - D(x - xw) (6.23a)
and
ds 1
- = -11(S)X + D(so - s) (6.23b)
dt YX1S
The steady-state values of this model as a function of D are shown in Fig. 6.6.
Note that the biomass concentration does not vanish above the washout point
(see also Sinclair and Brown, 1970; Toda and Dunn, 1982). Imperfectly mixed
bioreactor systems have been summarized recently by A. Moser (1985b).
6.1.2.3 Product Formation

The microbial product-formation rate in a single CSTR is (Pirt, 1975)
dp
dt = qp' x - D . x (6.24)
The change of product concentration in the culture over time depends on the
term of synthesis (qp' x) and on the term of outflow of product ( - D x).
In the stationary state with dp/dt = 0
_ qp.5(
p=-- (6.25)
D
where p and 5( are the steady-state concentrations of product and biomass,
respectively. If the product is strictly "growth linked" (cf. Equ. 5.116), the
product concentration in the stationary state p and the output rate D P will
__ 1~----------~ x,p 1 ,:---,----------.,. 1 D.p

X,p
D.p

a b
FIGURE 6.7. Growth-associated (a) and non-growth-associated (b) product formation,

growth, and productivity D . P in CSTR operation without P inhibition. YP1X = 0.5 and
kp = 0.1.
vary with D in the same manner as does biomass concentration (cf. Fig. 6.1).
If the qp is independent of the growth rate (cf. Equ. 5.120), the product
concentration varies inversely with the dilution rate D. Then, over a wide
range of dilution rates, the output rate is constant, as shown in Fig. 6.7a.
However, as D tends to zero, eventually the assumption that kp = constant
becomes invalid because of the decay of enzyme activity, because some
required substrate will be exhausted, or because of regulatory processes in the
cell. If the product formation is partly "growth linked" and partly independent
of the growth rate (cf. Equ. 5.122), then the product concentration in the steady
state will vary with dilution rate D, as shown in Fig. 6.7b.
The product output rate, if D tends toward zero, is the "non-growth-linked"
contribution k p , multiplied by biomass concentration x. From Equ. 6.25 we
get
D'p = qp'X (6.26)
Substituting the logistic form of product formation kinetics (see Equ. 5.122)
for qp gives
D P = (Yp!x' f1 + kp)x (6.27)
Rearranging with x= YX!s' So yields
D . P = (Yp!x' f1 + kp) YX!S . So (6.28)
and the product output rate, in the case when D tends toward zero, becomes
D . p = YX!s' So . kp (6.29)
The kinetic aspects of product formation in continuous culture are usually
dealt with using the simple, formal kinetic approach of the Luedeking- Piret
equation (Equ. 5.122). A modified kinetic approach to microbial product
formation uses the concept that rp not only is proportional to the actual
biomass concentration but also depends on a second carbon source Sz
(Hegewald et aI., 1978):
- _ o.4,,------~---:::I 1 51 ,S
X,p
A
1_0
FIGURE 6.8. Prediction of steady-state values of CSTR operation from batch data using
two substrates for growth and turimycin production with Streptomyces hygroscopicus
on the basis of modeling the kinetics with the aid of Equ. 6.30 (Hegewald et al., 1978).
S2
rp = qp,max +K . x . irepr - D .P (6.30)
S2 S2
This equation was later extended to supplementary O 2 limitation (Bajpaj and

Reuss, 1980). The factor irepr in Equ. 6.30 represents the effect of easily
metabolized sugars in reducing antibiotic synthesis via repression or inhibi-
tion (cf. Equ. 5.106). Figure 6.8 shows the calculated prediction of CSTR
behavior based on Equ. 6.30.
6.1.2.4 Product Inhibition

The difference between batch and continuous culture techniques with respect
to the effect of product inhibition is that under the conditions of continuous
cultures product is diluted, while in batch runs product is accumulated.
Therefore, in batch cultures reaction rates eventually slow down. In chemostat
cultures, however, oscillations of x and p appear due to periodic effect of, for
example, pH control and/or permanent inflow and outflow of fresh medium.
For the mathematical modeling and computer simulation of this problem, it
is possible to formulate the following differential equations:
dx
dt = /l(s,p)'x - D x (6.31a)
ds 1
-d = D(so - s) - /l(s,p)x'- (6.31b)
t Yx1s
dp
dt = qp' x - D p (6.31c)
Extensive studies of oscillations induced by products are given by Knorre

(1980) together with mathematical modeling and the problem of multiple
steady states (see next section).
Inserting a generalized kinetic equation for pure product inhibition from
the previous chapter (Equ. 5.63 with Equ. 5.105) into the basic performance
equation of a single CSTR (cf. Equ. 3.90) and converting all concentrations
I
I
!, SUMITlNG
iI
!
----------------- ---;, -- -- ------ -- - - ---- ---- --------
PMAXIMUM
Perit
/ i LARGEST SLOPE P
I :r /'
INITIAL / I P'M~>Y
SLOPE / : .<:
POPTIMUM ,
---------------4---
r
/..-: : ' OPTIMUM OPERATION
I
I
//, :
,
I
I
///// I
I
I
O~-----:1(---_--.i;;2--------PmaxT
WASHOUT tOPTIMUM
FIGURE 6.9. Product inhibition kinetics in a CSTR with Xo = 0, demonstrated in a plot

of concentration versus time (mean residence time t) cf. Fig. 6.1 b, (Levenspiel, 1980).
IgT
I
FIGURE 6.10. Method for evaluating the rate con-
I
stants for product inhibition kinetics from CSTR I
experiments with Xo = 0, according to Equ. 6.33 Igl-lmax /
(Levenspiel, 1980).
into P (Levenspiel, 1980 and Han and Levenspiel, 1987), gives for the case of
Xo = 0
Jlmax' t = (1 _~)-n Penl

for Jlmax . t> 1 (6.32)
The properties of Equ. 6.32 are displayed in Fig. 6.9 (cf. Fig. 6.1 b), which shows
that washout occurs at Jlmax t = 1 and that the maximum production rate is
dependent in a simple manner on Peril and n.
To find the kinetic constants from CSTR experiments, first evaluate Peril in
a batch run using an excess of substrate and letting t -+ 00. Then rearrange
Equ. 6.32 to give
log t = -log Jlmax + n log Peril (6.33)
Peril - P
and plot as in Fig. 6.10. The slope and intercept of the best line through the
data will then give the kinetic parameters /1max and n in the case of pure product
inhibition without substrate dependence. In situations where both substrate
availability and product inhibition affect the rate, the complete rate equation
(cf. Equ. 5.105) must be used, as previously outlined in Figs. 5.lOb and 5.33.
This is due to Equ. 6.34
~ 1 Ks 1
t=-+-~ (6.34)
kobs kobs S
6.1.2.5 Substrate Inhibition

The performance of CSTR is drastically changed by substrate inhibition
kinetics. All systems previously presented have the property of equifinality,
with only one unique steady state, and the attraction domain contains all
states. But in nonlinear open systems, additional steady states and additional
attraction domains are possible. A CSTR with S inhibition is an example
where there are two stable steady states (1 X, 1 sand 2X , 2S) and where the phase
plane is dissected in two nonoverlapping attraction domains.
The added factor 1/(1 + s/K 1 S ) in Equ. 5.88 represents the toxicity of the
substrate at higher concentrations. Let us recall that the condition for calcula-
tion of the stationary state with non vanishing biomass concentration is the
relation /1(s) = D. This equation has only one solution if /1(s) is a monotonic
function. But with characteristics as in Equ. 5.88, there are two solutions.
Together with the washout state ex, IS) we have three stationary states. Two
of them are stable ex,IS and 2X , 2S), one of them is unstable ex, 3 S ). Thus, we
have a bistable system. The stationary values of the stable and the unstable
stationary state are shown as a function of D in Fig. 6.11. Hysteresis may occur
in shift experiments. Figure 6.12 shows how the final biomass concentration
depends on the initial concentration. Figure 6.13 demonstrates that the phase
plane is divided into two attraction domains. Both domains are touched by
a separatrix in which the unstable stationary state lies. Note that, after an
external disturbance, the system can cross over the separatrix and shift from
one steady state to the other. This bistable behavior is a serious problem in,
for example, waste treatment: It takes place if substrates such as alcohols,
phenols, or hydrocarbons occur in such high concentrations that the utiliza-
tion of these substrates is inhibited.
6.1.3 STABILITY ANALYSIS AND TRANSIENT BEHAVIOR OF

THE CSTR
6.1.3.1 Stability Analysis

The steady-state behavior of a single CSTR can be calculated using Equ.
6.3a, b, representing the long-term behavior of the model system. However,
nonlinear algebraic equation systems may have more than one solution. To
- - - ,,
x,s ,,
,
A ,
,
.5
t
FIGURE 6.11. Effect of substrate inhibition on the behavior ofCSTR: For Db < D < Dc,
two steady states ex,
IS) and ex,
2S) and an unstable steady state ex,
3 S) coexist. The
dotted line sketches the hysteretic behavior for a shift-up/shift-down cycle where Db
and Dc have been passed.
)(
(91 'J
A
a -- s 1(9[')
.5
FIGURE 6.12. Bistable growth kinetics for growth limitation by an inhibiting substrate
in a cit plot. The inserted graph shows the corresponding kinetics. (Cf. Fig. 5.33.)
.5
----------------=:::::::~~~lJ ('5,')1)
o 1..1-_ _ _ _ _ _ _ _- ' -_ _ _ _ _ _ _ _--'-'
o .5 1(gr')
FIGURE 6.13. Bistability for growth limitation of an inhibiting substrate shown in a

phase-plane diagram. The attraction domains of the two steady states are touched by
a separatrix.
denote the various solutions we add an upper index to the symbols of steady-
state concentrations (x, s). The system of Equs. 6.6 and 6.7 has two solutions
(Gutke, 1980). The first solution we call the "washout state," because it is
characterized by a vanishing biomass concentration:
(6.35)
The second stationary solution is characterized by a non vanishing biomass
concentration:
2- 2- ({ Ks . D } Ks D ) (6.36)
( x, s) = YxlS So - Yxis _ D' _ D
Ilmax Ilmax
Both solutions are included in Figs. 6.1a, 6.4, 6.5, and 6.1l.
Starting the second step of analyzing the long-term behavior we may ask
these questions: Are there two stationary states or is only one of them realiz-
able in a practical case? What kind of changes occur in the system ifthe system
in one of these states is disturbed? These questions can be answered by
application of the stability theory, using analytical or numerical calculations
or computer simulation. A stationary state is called "stable" if the system
returns to this state after any kind of disturbance. Such a stable stationary
state is called a "steady state." For our model one can demonstrate that the
washout state (1 x, 1s) is stable for dilution rates greater than a critical dilution
rate Dc. The critical dilution rate is called "washout point": For dilution rates
below this washout point, the washout state is unstable, and only a stationary
state with nonvanishing biomass concentration is stable.
In contrast to closed discontinuous (batch) cultivation systems, in which
the final concentrations depend on the initial concentrations, in an open
system in continuous culture the final concentration is largely independently
of the initial states, which phenomenon is called "equifinality." An appropriate
plot for demonstration of equifinality is the plot of so-called trajectories in the
phase plane, as shown in the next section and in Fig. 6.13.
General considerations concerning stability analysis of biochemical reac-
tion systems have been presented by Stucki (1978). The problem of mono-
stability, bistability, and multiplicity (multiple steady state) is further discussed
in review articles and books (Aris and Humphrey, 1977; Bergter, 1972; Chi et
aI., 1974; Fiechter, 1982; Gutke and Knorre, 1980; Knorre, 1980; Koga and
Humphrey, 1967; Prokop, 1978; Romanovsky et aI., 1974; Russell and Tanner,
1978; Yang and Humphrey, 1975; Yano and Koga, 1973).
6.1.3.2 Transient Behavior of the CSTR

Apparently simple models are not always suitable for the description of the
dynamics ofbioreactor operations. As shown in Fig. 6.14, there are six classes
of transient behavior of the CSTR:
- I x monotonically decreasing, s monotonically increasing
- II x monotonically decreasing, s undershooting
- III x overshooting, s undershooting
- IV x monotonically increasing, s monotonically decreasing
- V x monotonically increasing, s overshooting
- VI x undershooting, s overshooting
Figure 6.14 shows for all of these six classes one representative example in the
plot of biomass (a) and substrate concentration versus time (b).
As previously outlined, a plot of trajectories in a phase plane x versus s is
advantageous: In such a graph the time factor is eliminated. The sequence of
states [x(t}, s(t)], which has passed from the initial state [x(O), s(O)J to the
steady state (X,8), is called the "trajectory." Figure 6.15 shows trajectories for
various initial states. All trajectories tend to the same steady state ex,
28). But
their transient behavior is different.
In considering the transient behavior of a CSTR after a shift in the dilution
rate, shift-up and shift-down experiments must be distinguished (see Fiechter,
1982). Assuming that the culture before the shift is in a steady state (mathe-
matically defined by Equ. 6.35 with D = Dd, consider the shift from Dl to D2 :
According to Equ. 6.35 (Fig. 6.1a), after a shift-up the biomass concentration
must tend to a smaller value, but after a shift-down it must converge to a
higher value. From Equs. 6.4 and 6.5, it follows that the balance relation
between the biomass and the substrate concentration must be maintained
.7
}(
(g./-'J
~
~I
~ __-=====__________x__
N
.2
.7
.ll
.2
.7
.2
o
>_;---====--------5--
a 10 -t 20 h
5
5.
(gt
o
.5
o
~
5\~
~====--- __------s---
0u-_ _ _~_ __ L_ _ _ _~_ _ _ _~_ _ _ _ _ __u
b o 10 _t 20 h
FIGURE 6.14. Transient behavior of biomass (a) and substrate concentration (b) from
different initial states to the steady state using Monod kinetics with six classes of
transients (I - VI).
FIGURE 6.15. Trajectories in the r.

phase plane for Monod kinetics
i
I
with the six classes of transient
I
:
behavior from Fig. 6.14. ~
III
~
II I
I..... I
l~
I:"--~
~ I
-
I '~
VI I
I ~
............ IV
I
I V '~
"'...
..........
I ' .....
o
It,S
(gn . 0-0.-' .... .0.75
II ..
.0.-'
.
A
.75
\
.5
x \
p \ ..:
o
5
( \ o
o _t tOOh
FIGURE 6.16. Monotonic transient behavior after shift-up and shift-down according to
Monod kinetics, using parameter values as in Fig. 6.4.
after the shift. Thus, both steady states must lie on the straight line x = YX1S
(so - s). This line lies within the domains I and IV of the phase plane shown
in Fig. 6.15. Thus, the response after the shift in the dilution rate is expected
to be a smooth transient to the new steady state. This prediction on the basis
of the Monod model is demonstrated by computer simulation in Fig. 6.16.
But this behavior is in contrast to that found experimentally: In experiments,

an overshoot of substrate concentration and of specific growth rate are
observed after a shift-up. After a shift-down the transient to the new steady
state takes place faster and monotonically. The biological interpretation for
this behavior is that in the case of shift-up the RNA concentration must
increase to increase the capacity for protein synthesis appropriate to the
higher supply of substrate. Thus, the specific growth rate follows the increase
of substrate concentration with a time lag (td. Conversely, in the case of
shift-down the substrate supply is decreased and the capacity for protein
synthesis cannot be fully employed. Thus, the specific growth rate follows
the decreasing substrate concentration without any lag. For discontinuous
culture, we introduced a lag factor in a simple way in Equ. 5.72, and in a more
generalized way by using Equ. 5.149. For tL = 0 we return to the direct,
unretarded Monod equation (Equ. 5.38). To formulate this we have a lag for
shift-up but not for shift-down, and we write the following equations:
set) .
pes, t) = Pmax Ks + set) - tL . P for {l > 0 (6.37a)
set)
pes, t) = Pmax KS + set) for {l :s; 0 (6.37b)
Note that this more general formulation of time lag is also suitable for
modeling a lag in a discontinuous culture. Figure 6.17 shows the transient
X,S D.O,t 0.75

(g'(') ....
.75
.5
5
a '-------J 0
o -t IOOh
FIGURE 6.17. Nonmonotonic transient after shift-up and monotonic transient after
shift-down according to Monod kinetics, modified by a lag time tL'
6.2 Variable Volume CSTR Operation 325
behavior of the modified Monod model where Equ. 5.38 is replaced by Equ.
6.37a,b. The simulation shows an overshoot of sand J1. after the shift-up,
whereas after the shift-down we have the same monotonic response as shown
in Fig. 6.16. Thus, the modification of Monod's model with Equ. 6.37a,b is
suitable for simulating shift experiments.
6.2 Variable Volume CSTR Operation (Fed-Batch

and Transient Reactor Operation)
Although continuous culture methods are commonly associated with phys-

iological studies of microbial populations in the steady state, these methods
have considerable advantage in studying behavior under transient conditions.
Examination of transient behavior in a continuous culture provides valuable
insight into the mechanisms of regulation in biological systems (Harrison and
Topiwala, 1974). The effects of a continuously changing environment are
significant in both natural and industrial microbiological systems. Without
resorting to strain selection, medium optimization, or new reactor designs, a
significant increase in product yield can sometimes be obtained by a very
simple operational mode: transient reactor operation (Pickett et aI., 1979a).
Process improvement obtained with periodic operation has been shown to
depend on the reaction rate constants. In the case of consecutive competing
chemical reactions (e.g., kl' k2' k 3 ), no yield or selectively improvement occurs
if kl k2 or kl k 2. With parallel reactions, a 20% increase over the steady-
state operation has been observed (Dorawala and Douglas, 1971).
Although biological reactors generally operate in a batch or sometimes in
continuous mode, some fermentations, termed "fed-batch cultures" (Yoshida
et aI., 1973), utilize a continuous periodic operation technique. This transient
condition takes several forms. In "repeated" fed-batch culture (Pirt, 1974),
complete medium is fed continuously to a batch culture; the resultant biomass
is reduced by being pumped out at preset intervals, and the process is then
repeated. In fed-batch culture, the dilution rate is therefore continually chang-
ing and represents the transient of the system. Some processes utilize a feed
of the (limiting) carbon source only; this type of fed-batch culture has been
termed "semibatch" culture (Yamane and Hirano, 1977). A further modifica-
tion of fed-batch culture, called "extended" culture, has also been reported
(Edwards et aI., 1970). This technique is more controlled than other types of
the fed-batch culture, since environmental sensors directly measure specific
parameters of culture growth and regulate the addition of the limiting carbon
source to the culture. Both semibatch and extended-batch types of culture
usually only cover one fed-batch "cycle" and so differ from repeated fed-batch
culture in the duration of the periodicity applied to the culture (see Figs.
3.31-3.33).
For quantification of variable-volume CSTR operations, see Equ. 3.86a,b
and the performance equation derived for a variable-volume CSTR operation,

Equ.3.92.
Mathematically, in terms of volumetric flow rates, liquid densities p, and
volume
dey pd
dt = Fin' Pin - Fex . Pex (6.38)
Normally, the density changes are small, giving

dV
dt = Fin - Fex (6.39a)
The total mass balance gives the rate of change of volume with inlet and outlet
flow rates for a well-mixed constant-density system. A fed batch is a special
case of a variable-volume CSTR operation: It has been defined as a bioreactor
with inflowing substrate but without outflow. For this system, the equation
becomes
dV
dt = Fin (6.39b)
where Fin can be a function of time F(t), according to Equ. 3.86. Equation
6.39b together with Equ. 3.92 provides the mathematical model for the fed-
batch reactor, and the behavior calculated by computer simulation is shown
in Fig. 3.37.
The variable-volume system considered here has three dependent variables:
V, XeX' and Sex' The equations previously mentioned provide the required
number of independent equations: Their simultaneous solution will yield the
variation of V, XeX' and Sex with time. The assumptions and restrictions are
as follows: (a) well-mixed reactor, (b) single limiting substrate, (c) balanced
growth, (d) constant yield coefficient, and (e) Monod kinetics. The restriction
that balanced cell growth must exist for the Monod relation to be valid puts
a strong limitation on the possible rates of change that can be considered. For
example, large differences in dilution rates would lead to rapid changes in
substrate concentration, and the specific growth rate could not possibly adapt
according to the Monod equation.
It is instructive to compare the fed-batch model with the model for a
constant-volume chemostat (see Equs. 3.90 and 3.91). Equation 6.3a,b is
formally identical with Equ. 3.92a,b, but it is important to note that the
physical meaning of the terms is not identical. Comparing the term - Fex .
xex/V in Equs. 3.92 and 6.3 it can be seen that the origin of this term for the
fed batch is the expression X ex ' dV/dt; it thus represents a decrease in cell
concentration due to the volume change that arises from inlet flow rate Fin'
On the other hand, - Fin' Xex in Equ. 6.3, the chemostat, is a washout term
expressing the mass flow rate of cells that leave with the outgoing stream. A
fed batch can be compared with a constant-volume chemostat whose feed rate
is decreasing slowly.
6.2 Variable Volume CSTR Operation 327
Because the mathematical form of the fed-batch model Equ. 3.92a,b is

identical to that of the chemostat Equ. 6.3a,b, it can be concluded that the fed
batch will behave analogously. A dynamic steady state will be achieved for
sufficiently low flow rates such that the specific growth rate is maintained
exactly equal to the dilution rate FinlV This phenomenon has been identified
previously, and the dynamic steady state has been termed a "quasi-steady
state." It is characterized by a constant value of X ex , which must exist because
p, = FinlV Since the volume is increasing steadily, p, must be maintained by a
decrease in Sex; therefore, dsexldt is not zero. Computer simulations obtained
by solving Equ. 3.92a,b together with Equ. 6.39b numerically show that the
phenomena do indeed occur as described. The computer simulation shown in
Fig. 3.37 indicates that substrate concentration steadily decreases during the
quasi-steady state p, = D, in order to maintain the quasi-steady state as the
volume increases (Dunn and Mor, 1975). When the quasi-steady state is
achieved, the equality of p, and FinlVleads to the following relationship for Sex:
Fin Ks
S =-.----- (6.40)
ex V P,max - (FinIV)
For the special case of constant feed rate, Fin> the fed batch, has further
properties that are worth noting:
1. From Equ. 3.92a, under the quasi-steady-state conditions, p, = FinlV, and
it follows that
d(V xex)
--d-t---'- = Fin Xex = constant (6.41)
The rate of change of total biomass is constant.

2. From Equ. 6.39b, the volume must increase linearly with time:
V = Vo + Fint (6.42)
3. The rate of change of p, during a quasi-steady-state period is
d Fin _ Fi~ _ Fi~

(6.43)
V - - V2 - - (Vo + Fin tf
At low t when V = Vo this becomes
dp,
(6.44)
dt
at higher values of t or conditions when V Vo
(6.45)
Thus, it is seen that p, decreases most rapidly during the first period of the fed
batch, the rate of decrease becoming slower with time. For the purpose of
obtaining a particular desired rate of change of p" the flow rate Fin could be
changed progressively during the course of fermentation (Dunn and Mor,

1975).
Fed-batch operations are of great theoretical and practical importance. For
basic research in process kinetics, a fed-batch process is advantageous-the
extended culture offers the possibility of maintaining low S concentrations
over long periods of time, and this facilitates measurement of kinetic param-
eters (Esener et aI., 1981c; Lee and Yan, 1981). Several computer simulations
have been carried out to demonstrate fed-batch productivities compared with
the CSTR at constant volume (Keller and Dunn, 1978a) and to show the
influence of such factors as fluctuations in volume or S feed (Keller and Dunn,
1978b). Computational results also provide quantitative sensitivity analysis
that is very useful in determining the degree of precision necessary in applying
process equations (Kishimoto et aI., 1976). Furthermore, the fed-batch process
provides the same information as the CSTR without the requirements for an
outflow, without volume control, and without the necessity of shifting steady
states. Bioprocessing on a larger scale profits from the fact that fed-batch
cultures show the unique feature of transient conditions, with the growth rate
under control between fixed values. There is evidence that the maximum rates
of some processes can be achieved only transiently: Examples are productions
of some bacterial antigens (Pirt et aI., 1961), synthesis of beta-galactosidase
(Knorre, 1968), biosynthesis of secondary metabolites such as penicillin
(Bajpaj and Reuss 1981; Bajpaj and Reuss, 1980; Court and Pirt, 1976; Heijnen
et aI., 1979; Pirt, 1974; Wright and Calam, 1968), turimycin (Gutke and Knorre,
1981), streptomycin (Hegewald et aI., 1978), and cephalosporin (Matsumura
et aI., 1981; Trilli et aI., 1977). A typical and famous example of a drastic
increase in productivity in case of secondary metabolite production is shown
in Fig. 6.18 by choosing the adequate level of S dosing and OTR capacity
(Bajpaj and Reuss, 1980, 1981). In addition to these case involving catabolite
repression and/or inhibition (Demain et aI., 1979; Gutke and Knorre, 1982),
fed batch processes are advantageous in the cases with complex kinetics in
yeast technology (Aiba et aI., 1976; Dairaku et aI., 1981) and in waste water
treatment in which substrate inhibition is also commonly encountered. The
increase in cell mass, derived from Equ. 3.43, for the quasi-steady state is given
by
dx
V- = p. ys (6.46)
dt 0
and the cell mass to be harvested at this condition is calculated by

x V = Xo . V + p. y. So . t (6.47)
Transient reactor operations have been summarized in review articles (e.g.,
Barford et aI., 1982; Douglas, 1972; Pickett et aI., 1979a; Pickett, 1982) and
appear increasingly in the literature (e.g., Borzani et aI., 1976; Chi and Howell,
1976; Cooney and Wang, 1976; Daigger and Grady, 1982; Klei et aI., 1975;
Parulekar and Lim, 1985; Pickett et aI., 1980; Regan et aI., 1971; Sherrard and
Lawrence, 1975; Sundstrom et aI., 1976; Yamane and Shimizu, 1984).
6.3 Multistage Single and Multistream Continuous Reactor Operation 329
Pmax
--- k L1 a -200 h- 1
- k L1 a 400 h-1
,
", \
"\
1\
I ,
I \
II '\
I
I
o 0.2 0.4
FIGURE 6.18. Computer simulation of product concentration Pmax (units penicillin/ml)

as a function of substrate feeding rate Fs [g '1- 1 initial volume' h -1] with varied kL1 a
value and initial substrate concentration So. At high values of Fs, productivity decreases
due to catabolite repression and O 2 limitation. (From Reuss et aI., 1980.)
6.3 Multistage Single and Multistream Continuous

Reactor Operation
The complexity of multistage systems increases with the number of stages. In
practice, however, it is rarely necessary to use more than three stages, and the
essential principles of multistage operation may be illustrated by a discussion
of only two stages.
6.3.1 CLASSIFICATION
Herbert (1964) distinguished three main types of multistage systems, as shown
in Fig. 6.19:
- Single-stream multistage (Fig. 6.19a)
- Multistream multistage (Fig. 6.19b)
- Multistage systems with recycle (Fig. 6.19c)
Single-stream systems are a cascade with a single medium inflow that is
constant through all other stages. The characteristic features are
- Later stages cannot affect earlier stages
- The first stage behaves as a CSTR
- Dilution rate D cannot be changed in one stage without changing it in all
others, because the flow is constant
- The dilution rate in individual stages depends only on the reactor volume
N-1 N-2 N-3
5
F__ ~+-~ __~~__+-~~ ~~~~.,--.F
FIGURE 6.19. Different types of multistage CSTR operation: (a) single-stream multi-
stage system, (b) multistream multistage system, (c) multistage with recycle.
(6.48)
Multistream systems have multiple inputs of feed and therefore are charac-
terized as follows:
- Earlier stages are again independent of later stages, and the first stage
behaves as a CSTR
- The different medium feeds may be varied independently and individual
dilution rates are independent process variables
6.3.2 POTENTIALITIES OF MULTISTAGE SYSTEMS

Multistage systems have seldom been used in industry or research. From a
theoretical viewpoint, the potential advantages are quite promising (Ricica,
1969a). In general, multistage systems provide different environments in each
stage and thereby approximate the behavior of tubular reactors (cf. Fig. 3.30).
Multistage systems, therefore, potentially offer the advantages of CPFRs (cf.
Sect. 6.4), which can be summarized as follows:
- Maximum conversion is achieved by complete utilization of the substrate
in the later stages and maximum productivity in the first stage. This is
essential in the case of expensive substrates, for example, steroid transfor-
mation, or in the case of environmental problems, for example, waste water
treatment.
- All physiological states of microbial cultures are accessible, including the
"mature cell" stage necessary for some types of product formation such as
secondary metabolities.
- Long residence times can be maintained. This is advantageous in the case
ofbioprocesses in the stationary growth phase with complex kinetics or with
complex media.
- A certain product quality may be desired (e.g., bakers' yeast), and multistage
processing for this purpose is already found in industry.
- Optimal environmental operating conditions (for temperature, pH, and
Po) can be maintained in such gradient reactors.
6.3.3 SINGLE-STREAM MULTISTAGE OPERATION
Consider a two-stage reactor system in which the first stage behaves like a
single CSTR (Fig. 6.19a; Herbert, 1964; Malek and Fencl, 1966; Ricica and
Necinova, 1967). The second stage may be treated in much the same way as
the first, but the equations will be more complicated, since x and s enter the
second stage. The balance equations for x and s are
dX2
- = D2 'X 1 - D 2 'X 2 + 112' X 2 (6.49)
dt
and
dS 2 1
-d =D2'Sl-D2'S2--'112'X2 (6.50)
t YX1S
where 112 follows simple Monod kinetics. In the steady state with dx 2 /dt = 0
(6.51)
and with ds 2 /dt = 0
112 = Y' D 2 C~S2)

1 (6.52)
Eliminating 112 from Equs. 6.51 and 6.52 gives

X2 = Y(so - S2) (6.53)
from which X 2 can be calculated when S2 is known. To find 52 we have, from
Equ. 6.52 and Monod kinetics by eliminating 112,
(6.54)
and eliminating X2 from Equs. 6.54 and 6.53 gives

YD 2
Y(so - S2) = (Ks + S2)(S1 - S2) (6.55)
f1.max S2
Solving this equation for S2 after substituting Equ. 6.6 gives a quadratic
equation
+ K s D1 D2=0
2
(6.56)
f1.max - D1
in which S2 is expressed as a function of So, D1 , D2 , and f1.max, Ks. The quadratic
equation has two positive roots and gives two values for S2' One smaller and
One greater than S1; the first is the correct value, and the second corresponds
to the biologically imaginary but mathematically real case of "reverse" growth
(with conversion of cells to substrate). Thus, from Equs. 6.53 and 6.56 the
steady-state concentration of cell and substrate in the second stage can be
calculated, and the solution is shown graphically in Fig. 6.20. Extension of
this mathematical analysis to three or more stages can be continued, the
equations for each fermenter being derived from those of the preceding One.
For the Nth reactor in a cascade, the material balance equations using
symbols shown in Fig. 6.19a are, therefore, given by
(6.57)
(6.58)
(6.59)
x,s
X
...1,-,\ FIGURE 6.20. Theoretical relationships be-
,,/"'O~ \ tween stationary values of the cell mass con-
,;;(" \ centrations Xl and x2 (or substrate concentra-
~~/ \ tions Sl and S2) and the dilution rate in a
/~~ OX2\ S1 . / S2 cascade with N = 2. The productivities are
..~-:.....::..:..:..'-----0 designated Dx. (Adapted from Herbert, 1964.)
""------'-----'-:.:.-::...
Thus, at steady-state conditions

D'XN - 1
XN = ----'- (N"# 1) (6.60)
D-J.l.N
1 YPIX
SN = SN-I - -D v-' J.l.N . x N (6.61)
IXIS
1
PN = Jj(D . PN-I + Yplx' J.l.N XN) (6.62)
The number of stages N and the parameters of interest for a multistage system
can be determined by a graphical method described by Deindoerfer and
Humphrey (1959) or Luedeking and Piret (1959) in analogy to the chemical
process design called the "periodic equilibrium curve." This method enables
the calculation of initial estimates and forms the hypothetical model (cf. Fig.
2.18). In application, kinetic data from measurements in a DCSTR are plotted
as rx against x (Fig. 6.21). The productivity is always a combination of kinetic
and mass conservation elements, and for the Nth reactor in a continuous
process, from Equ. 3.89 (cf. Equ. 6.57)
dX N F
Tt = -V(XN - xN-d (6.63)
which is entered in Fig. 6.21.

For the first stage, with xo.c the straight line represented by Equ. 6.63 with
the slope F /V! = DI = J.l.I is drawn. The intercept of the line with the kinetic
curve (labeled point 1) gives the concentration x for N = 1. An optimal reactor
MASS BALANCE FOR NCSTR
KINETICS
DISCONTINUOUS PROCESS
" D :
_ "'r' 2 _I X
x a,e Xa,de
FIGURE 6.21. Graphical procedure for obtaining the dilution rate through the system,
D, for a cascade of stirred vessels (NCSTR) in carrying out a continuous microbial
growth process on the basis of discontinuous experimental data. The discontinuous
process kinetics in the rxlx plot are analogous to a plot from a chemical process.
system would operate with the maximum slope: Dl """""* I1max' However, due to
the danger of washing cells out of the reactor, the actual optimum, Dop " is
selected according to Fig. 6.1a (cf. Equ. 6.11). The first intercept, the one at
the lower region of the curve, represents an unstable operating point (the lag
phase).
In the same way, additional curves can be drawn for N = 2,3 ... (Equ. 6.63),
each with the starting concentration of the previous reactor in the series. The
slope of the straight line represents the value of D for each case. When F is a
constant for the whole system, the volume of each stage of the cascade may
be optimized.
In developing a process with a continuous cascade of stirred vessels, it
should be remembered that the productivity (in terms of product per unit
volume per unit time) is not optimized. Rather, the conversion (in terms of
substrate utilization) is maximized. To clarify this, Fig. 6.20 (similar to Fig.
6.1a) is the diagram of a two-stage process (Herbert, 1964). One sees here
that the productivity of the two-stage CSTR (Pr = D x2 ) is less than D Xl'
but the use of substrate (S2) is more complete than with a single stage CSTR
(Sl)'
In considering the whole system it is useful to calculate its average rate,
Daw defined as the total flow through the system divided by the total volume
of culture in all fermenters. For a single-stream chain of N fermenters this is
F
Dave =--
N. V (6.64)
The washout rate for the whole system, that is, the critical value of Dave, is
therefore
(6.65)
in a third or subsequent fermenter. In fact, if calculations for a third and fourth

stage are made, the resulting curves are scarcely distinguishable from curve
X 2 in Fig. 6.20. This suggests that there will seldom be much practical advan-
tage in using more than two stages, at least for the quantity production of cells.
On the other hand, further stages might be important in obtaining cells of a
desired quality. Also, using a two-stage process, a continuous culture may be
maintained at a true1y maximum growth rate; this is impossible in a single-
stage process.
6.3.4 MULTISTREAM MULTISTAGE OPERATION
We may conclude from the previous section that the experimental possibilities
of single-stream continuous systems are rather limited. There are many more
possibilities with multistream systems, since these have multiple medium
inflows that are independently variable and that may, if desired, contain a
variety of different nutrients. This can produce very complicated experimental

situations. We will consider only the simple case in which a single growth-
limiting nutrient is used in a two-stage system. (Fig. 6.19b). Systems with
different substrates fed to later systems are referred to (Herbert, 1964).
The dilution rates of the two fermenters are respectively D1 = FdVl and
Dz = Fz/Vz . It is convenient mathematically to consider Dl as the sum of
two "partial dilution rates" DOl and D 12 , due to the two different inflows to
fermenter 2, and defined as
FOl
DOl = V;' (6.66)
Again, the first stage is described with Equs. 6.6 and 6.7.
For the second stage, the mass balance equations for x and S are
dXl
---;{t = D12 Xl - Dl . Xl + Ill' Xl (6.67)
and
at
dS l = D12 . Sl + DOl' SOl - D z ' Sl - Y ) Xl
(Ill
X1S
(6.68)
In the steady state, these equations become
(6.69)
and
Xl
III = y(D 12 s 1 + DOZs Ol - DZs l ) (6.70)
Eliminating Ill' Xl and rearranging gives

D11 . SOl DOl' SOl )
Xl = Y( + D - Sl (6.71)
Dl 2
from which Xl can be calculated when Sl is known. To find Sz on the basis of

Equ. 6.70 and the Monod relation for Ill' which is eliminated, we get Equ.
6.72 after again eliminating Xl and rearranging
- D12 'Sl + Ks' Dl)Sl + Ks' DOl 'SOl + Ks' D12 'Sl = 0 (6.72)
From this quadratic form, Sl can be computed by inserting the value of Sl

from Equ. 6.6, and the steady-state behavior for any value of D1 and Dl can
be derived; the only precondition concerns the values of kinetic parameters.
x
A
.-.-. _._-~._.- ._._--;/' /'
/'
/
5t-________x_1~______~./'~./___
./
~------------,_------------~--~o
200-2
FIGURE 6.22. Two-stream two-stage CSTR operation in the steady state: Growth rates
in first and second stage (/11' /12) and concentrations calculated with the help of Equs.
6.67 through 6.71 (Do2 = 1 h -1). (From Herbert, 1964.)
u u
A
FIGURE 6.23. Multigraphical interpretation of streptomycin production according to

Ricica (1969). The central graph represents the cit plot of batch data, with cit plots of
individual components in the upper graph (1, IJ(-amino nitrogen; 2, phosphorus; 3,
pyruvic acid; 4, pH; 5, ammonia nitrogen). The quantification is illustrated in the lower
part (/1, qs, and qp vs. t). The design of an NCSTR cascade with N = 6 is demonstrated
in the graphs on the left side (I and II) and right side (III-VI) using plots of llrj versus
cj (cf. Fig. 6.32.)
6.4 Continuous Plug Flow Reactors (CPFR) 337
The output for a single-stream chain of N vessels is

Fx N
Output = N. V = DaveXN (6.73)
The behavior of a two-stream, two-stage system is illustrated in Fig. 6.22.

In this case, the medium flow to the second fermenter is assumed to be held
at constant value while the inflow to the first fermenter is varied. The curves
of Fig. 6.22 are calculated for a constant second-stage medium inflow of
D02 = 1.0 and Dl values from zero to Dc (0.98). These curves illustrate very
clearly the stability of the system when operating at high second-stage growth
rates, and the relative insensitivity of the second stage to changes in the first.
It will be seen that alternations in the first-stage dilution rate over nearly the
whole of its working range (Dl = 0.15 to 0.9 h- 1 ) produce variations ofless
than 10% in the second-stage cell concentration and growth rate. The great
flexibility of multistream systems, combined with their stability and the ease
of operation at high flow rates, makes them superior to single-stream systems
in nearly all respects.
Figure 6.23 illustrates the design procedure in case of antibiotic production
in a six-stage cascade (Ricica, 1969a) using the design concept of Fig. 6.32.
6.4 Continuous Plug Flow Reactors (CPFR)
Reactors systems that have plug flow characteristics (that is, N ~ 5 or Bo ~ 7,

see Sect. 3.3) differ from CSTR types in that in the former case there is a narrow
residence time distribution. The whole mathematical theory of continuous
cultivation discussed so far has concerned the well-mixed stirred tank reactor.
6.4.1 PERFORMANCE EQUATIONS

An ideal case of a CPFR is represented by a narrow empty tube through which
the unstirred liquid flows uniformly (see Fig. 3.38). The concentration profile
was illustrated in Fig. 3.30. As washout is immediate, a constant inoculum or
a permanent recycle stream is required for continuous operation of a CPFR.
Since the composition of the fluid varies from position to position along the
longitudinal axis z, the material balance must be made on a differential
element of fluid dV = A . dz, as referred to in Sect. 3.4 (cf. Equs. 3.94-3.96).
Equations 3.95a,b and 3.96a,b form a set of nonlinear differential equations
that are difficult to solve analytically; they are more readily solved by com-
puter techniques. In special cases, however, the equations can be integrated,
for example, when the amount of biomass formed by the reaction is small
relative to the entering amount and the value of x is nearly constant over the
reaction length (xave). Thus, integration of Equ. 3.95a is possible, yielding the
following expression with Monod kinetics
( in Si _ Ilmax . Xave . A .Z
Si - s) + K s- s- - y. F(1 + r) (6.74)
The small change of x along the reactor is approximated by (x - xJ =

Y(Si - s); substituting for (Si - s) from Equ. 6.74 gives Equ. 6.75:
(x _ XI.) = x ave A z _ K . Yl ~
Ilmax (6.75)
F 1+r
() s nS
/ Km
/
/ --
r MAX
-K
Km+ So+ 2S 1Kis

k+2
FIGURE 6.24. Parameter estimation from integral reactors (DCSTR and CPFR) in the
case of simple enzyme kinetics (a), substrate inhibition (b), and competitive (c) and
noncompetitive product inhibition (d), according to Levenspiel (1979).
For enzyme processing, Fig. 6.24 summarizes some graphical methods for
parameter estimation from CPFR.
6.4.2 POTENTIAL ADVANTAGES OF CPFR OPERATION

Tubular reactors offer some potential advantages over conventional stirred
tank types of reactors (A. Moser, 1985b):
- High productivity and optimum conversion are realized simultaneously
- Mixing within tubular devices is more uniform eliminating dead spaces and
resulting in more secure scale-up (Russell et ai., 1974)
- Surface area-to-volume ratio is significantly higher resulting in facilatated
transfer processes. As a consequence, mass transfer is achieved with com-
parable less power consumption in the case of horizontal devices (Moser,
1985b; see also Kung and'Moser, 1986; Ziegler et ai., 1977) and heat transfer
is easy. to be realized. This fact becomes crucial in extreme situations of
bioprocessing e.g. with solid-substrates, photoreactions (maximum expo-
sure to light), shear sensitive tissues etc.
- Bioprocessing occur with gradients of concentrations and/or temperature
over the length of the tubes, so that adaption of technical manipulations to
biological demands can be carried out (e.g. T-programming as a function
of axial distance; substrate-dosing, COr removal etc., in the case of inhibi-
tion and/or repression kinetics (cf. Equ. 5.106). The case of product in-
hibition is quantified in Fig. 6.25. This property is advantageous also for
fundamental investigations of biological processes. Fig. 6.26 illustrates the
interrelations between biokinetics and optimal bioreactor design in case of
continuous operation (Moser, 1983b). Four cases of kinetics are shown
together with the bioreactor thought to realize highest conversion, which in
all cases is based on a CPFR-system.
- Horizontal versus vertical position is possible (tubular vs. tower reactors),
each having advantages. Horizontal configurations exhibit the property,
FIGURE 6.25. Demonstration of the ad-

vantage of CPFR over CSTR by calcu-
5
lating the volume ratio in dependence
of conversion reached (') as a function
of biokinetics using the simple Monod
equation and varying the ratio so/Ks
and also product inhibition kinetics with
variation of solK J A. Moser, 1985c, re-
printed with permission from Conser-
vation and Recycling, vol. 8, No. 1/2,
Pergamon Press, Oxford.)
KINETICS REACTOR
~ a

pCp)
st 'V-P
- cj
b
s
~ ---- --(P)
2 II
-P
C
't
I-I(S)
I
P
'1
It tt
....
' ........ I-I<P)
FIGURE 6.26. Graphic representation of the connex between biokinetics and corre-
sponding optimal continuous bioreactor design in four cases (a-d) (Moser, 1983b).
that plug- flow is not disturbed by the CO2 evolved and that hydrostatic
pressure cannot become inhibiting or cannot create practical problems.
- Practical advantages exist due to the closed system of tubes (no aerosol
formation in the case of waste treatment, easy operation under pressure
and/or with pure 02)' The technical apparatus is easy to handle because the
basic elements are commonplace (pipes, pumps, standard fittings). Also, wall
growth seems to be more easily controlled due to the high liquid velocities.
6.4.3 PRINCIPAL PROPERTIES AND DESIGN OF CPFRs COMPARED

WITH CSTRs
The CPFR has distinct advantages for those bioprocesses that call for a
specific reaction time, as illustrated in Fig. 6.27. Examples of such processes
include the formation of the products of secondary metabolism, sterilizations
to destroy cells and spores, and the high temperature treatment of foodstuffs.
f(t)
idCSTR
1 idCPFR
,
x
GIL
1.6
--:'=-=:-"\
..... '" , I \
I 4
,S
GIL
/' \
I
CPFR
"-
t< t t)t 0.8 2
CELLS WHICH DO NOT CELLS WHICH DO

CONTRIBUTE TO THE CONTRIBUTE TO
PRODUCTION IN PRODUCTION IN
CSTR CSTR
0.2 0.36 -0 h-1
6.27 6.28
FIGURE 6.27. Schematic comparison between an ideal CSTR and an ideal CPFR in
the case of a bioprocess maintaining a particular mean residence time t (production
of secondary metabolites with maturation time tM = t); sterilization and heat treatment
of foodstuffs with t = tSt '
FIGURE 6.28. Comparison of CSTR and CPFR operation in a steady-state diagram of

concentrations (x, ---; s, - - ) versus dilution rate, both operating under the same
recycle conditions (r = 0.4 and f3 = 3). (Adapted from Grieves et ai., 1964.)
6.4.3.1 Processes Involving Enzyme Kinetics

Levenspiel (1979) and Powell and Lowe (1964) compared a chain of CSTRs
with a tubular reactor by computing the biomass concentration obtained
in the final stage of a cascade of five chemostats (with feedback). It was
found that under certain circumstances a more complete utilization of growth-
limiting substrate can be obtained in the series of CSTRs than is possible with
the single CSTR. Similarly, Grieves et al. (1964) modeled a piston flow reactor
with recycle, compared the performance with that of a CSTR with recycle (cf.
Sect. 6.5), and concluded that if the objective of a process is cell mass produc-
tion, the differences are slight, but if the objective is reduction in effluent
substrate concentration, then a CPFR would be the optimum choice. With a
large recycle factor and with efficient operation of the separator resulting in
a large concentration factor, the differences in the critical residence time for
CPFR and CSTR become much less pronounced, and the CPFR is able to
provide a greater maximum production rate of microorganisms while yielding
a considerably lower effluent substrate concentration than a CSTR. This
behavior is illustrated in Fig. 6_28.
However, practical engineering experience with a tubular reactor, for exam-
ple, for biological waste water treatment (F. Moser, 1977, 1980) shows that
is,L (%)
100
y-.:;.~.----,:
"
60
t
CSTR
,
20
0,5 2 3 _ Bx KG COD.KG~dI
FIGURE 6.29. Conversion of substrate in the liquid phase ('s,d as a measure of

purification effect in waste water treatment versus mass loading rate per unit mass B.
in case of CSTR and CPFR operation: Compared are systems operating with one
common sedimentation tank (SED), exhibiting only a small difference in conversion,
and systems with separated SED, called "contact stabilization" or "sludge-reaeration,"
in which case the CPFR is superior to the CSTR. (Adapted from F. Moser, 1977.)
the CPFR is superior to a CSTR for S conversion only under certain condi-
tions (see Fig. 6.29). Comparisons were carried out using a single sedimenta-
tion vessel (case A) and using separated sludge settling devices (case B). The
theoretical advantage of a CPFR is experimentally realized only in case B, as
this example is complicated by the phenomenon ofbiosorption (cf. Sect. 5.3.9).
Biosorption does not play an essential role when fresh, unloaded sludge
is used after sludge stabilization, as indicated on the inserts of Fig. 6.29
(case B).
For attaining a desired conversion, Cex, one may calculate the necessary
mean residence time from
tOCSTR =
-t = -
fCe> -dc
1
== Co i~ex -dC
1
(6.77)
CPFR
Co r ~o r
For the ideal CSTR, the concentration is constant (cR = cex ) and
(6.78)
Equations 6.77 and 6.78 are generally applicable (for example, to sterilizations)
and can be formulated for use in the case of enzyme kinetics involving
substrate utilization as
-
t CPFR =
ISo> Y(Ks + s} ds (6.79)
So f-lmax s x
or
tCSTR
(so + Sex) Y(Ks + s}
= --"----=-::'-----'----- (6.80)
f-lmax s x
Using t = VjF, the volume necessary for a particular conversion in a CPFR
or CSTR can be calculated. The result of this type of calculation is shown in
Fig. 6.30 as the ratio of the volumes VCSTRjVCPFR as a function of the Ks value
(A. Moser et al., 1974). The various curves refer to differing values of So and
Sex> that is, to differing conversions. From this graph, the differing influences
of Ks may be seen. The Ks value determines whether a CPFR would be more
advantageous than a CSTR for any particular continuous bioprocess. The
range of numerical values for Ks is very low for enzyme reactions; even for
microbial growth at low concentrations Ks is approximately 10 mgjl (Monod,
1942). In the literature there is a large variation in the Ks values given for
bioprocesses from DCSTR and CSTR. The appearance of pseudo kinetic
parameters, especially Ks values, limits the reliability of predictions.
However, even at low Ks values, the CPFR continues to have a clear volume
advantage (as may be seen in Fig. 6.30) when the attainment of either a high
conversion or a low S concentration in the effiuent (sex ~ 20 mgjl) is an
important consideration (e.g., biological waste water treatment, (A. Moser,
1977; Wolfbauer, Klettner, and Moser, 1978). This consideration oflow Sex is
not of such great significance in fermentation processes (except perhaps with
expensive substrates), so the advantage of a CPFR is not nearly so pronounced
(Finn and Fiechter, 1979).
When one looks at real processes, one can see that they actually show
several different types of kinetic behavior. As shown in Fig. 6.31, reaction rates
(or conversion) vary as a function of concentration: There are normal catalytic
processes (curve a) and autocatalytic processes (curve b) that are dominant in
microbial growth (cf. Equ. 2.7) (Levenspiel, 1972). Biotechnological processes
are a combination of both (curve c). At the beginning they are autocatalytic;
later, after the exponential growth phase, they shift to ordinary kinetics.
When one recalls the equation used to develop the CPFR and CSTR, one
finds that it is possible to determine t graphically from a plot of 1jr versus c.
From Equ. 6.78, tCSTR is represented by the area of the rectangle with sides
equal to the final concentration, or the desired conversion. For t CPFR ' Equ.
6.77 applies, and the corresponding area is that under the curve. If one relies
on Fig. 6.31 and chooses a suitable representation of the kinetics of one of the
various processes (as shown in Fig. 6.32), t can be obtained. For a normal
catalytic process such as is found in enzyme technology, the CPFR will always
be advantageous (Lilly and Dunnill, 1971; Wandrey, Flaschel, and Schiigerl,
1979). For most biotechnological processes, the optimum configuration thus
5
4
3
---- B
2~~~~~;;;;~~~g~
c
50 100 200 Ks(mg/I) L---------------------__ c j
6.30 6.31
FIGURE 6.30. Calculated comparison of the reactor volume of a stirred vessel (VCSTR )
and a tube reactor (VCPFR) (cf. Equs. 6.77 and 6.78) demonstrating the influence of the
case So Sex (s
[gjl] [gjl] [%J
A 0.5 0.02 96
B 0.1 0.02 80
C 0.5 0.1 80
D 0.2 0.1 50
Ks value on the conversion of a continuous bioprocess with Monod kinetics (A. Moser
et ai., 1974).
FIGURE 6.31.Dependence of the reaction rate, the rate of formation or of consumption,
r i , on the concentration of component i, Ci , in (a) a normal catalytic process, for
example, an enzymatic process, (b) an autocatalytic process such as pure biological
growth, and (c) a biotechnological process such as fermentation or waste water treat-
ment with combined growth and product formation.
would be combination of a CSTR followed by a CPFR in a second stage

(Bischoff, 1966; Levenspiel, 1972; Topiwala, 1974; Yamane and Shimizu, 1982)
or a cascade of CSTRs (Ricica, 1969a,b).
The graphical method of obtaining t shown in Fig. 6.32 can be used, but t
may also be obtained with the aid of a mathematical representation of the
kinetics (cf. Equs. 6.79 and 6.80). Changing variables from the concentration
to the relative conversion (cf. Equ. 2.45) according to Topiwala (1974)
(6.81)
yields from Equ. 6.79
- f'"
t CPFR = _.
1 Ysod( (6.82)
f'" f'"
o rx
= Ks/so + (1 - () d( = 1(0 d(
o J1max(1 - 0 0
a
' - - - - - - - - - = C - - -...... c 1

end
FIGURE 6.32. Plot of the reciprocal of the reaction rate, 1/ri , as a function of the
concentration, Ci , as a measure of conversion. The processes shown in graphs a-c
correspond to Fig. 6.31a-c, which may be used in evaluating the mean residence time
for a continuous stirred vessel CSTR or a tubular reactor CPFR with Equs. 6.77 and
6.78. A combination of a CSTR with a CPFR is shown to be optimal in case c (stage
I is a CSTR and stage II is a CPFR). (Figures 6.32a and 6.32b adapted from Levenspiel,
1972; Figure 6.32c adapted from Topiwala, 1974).
and from Equ. 6.80

(6.83)
where
(6.84)
6.4.3.2 Fermentation Processes for Producing Secondary Metabolities

The kinetics of the production of secondary metabolites may be understood
with the help of the "maturation time" concept, Equ. 5.136. For a continuous
operation, the condition that t = tM should be maintained, that is, the mean
residence time spent by the suspended cells in the liquid phase, t, must be equal
to the ripening time, tM'
As a consequence of this condition, in a CSTR all cells with t < t(cells with
t < tM) are not yet ripe enough for production. Only those cells with a lifetime
t 2: t (t 2: tM) contribute to production. With the aid of the mathematical
function that describes the residence time distribution in the liquid phase of
a CSTR (Equ. 3.11), one can write
x(t) == x == f(t) = D e- D t (6.85)
The product concentration can then be written from Equ. 5.136 as
p= Yp!x' roo x(t) dt = Yp!x' X' e- D tM (6.86)

Jt M
In contrast, with a continuous tube reactor with t = tM , the maximum pro-

ductivity can simply and directly be calculated from
Prmax.CPFR = D P = D x' Yp!x (6.87)
A comparison of Equs. 6.86 and 6.87 easily shows that the productivity in a
CPFR is greater than that in a CSTR (cf. Fig. 6.27).
An additional advantage of the CPFR in this case comes from the fact that
some secondary metabolites (such as, for example, penicillin) are destroyed by
an excessive amount of time in the reactor. This was indicated in Equ. 5.128.
In a CSTR, all cells with t > tM and thus competent to produce the secondary
metabolite would at the same time be subject to the negative effect of the
destructive term, - kp,d'
6.4.3.3 Sterilization Processes

The destruction of microbial cells follows formal first-order kinetics for the
number of cells (cf. Equ. 5.267). The technical advantages of a CPFR process
over a CSTR process can again be seen in situations where t = tSt by con-
siderations analogous to those of Fig. 6.27.
The Tit profile of a continuous reactor must be taken into consideration in
any practical calculation for a sterilization, since kd is temperature dependent
(Equ. 5.3a). To determine the effective holding time at the sterilization tem-
perature, the following equation is a basic consideration:
(6.88)
The integral for the heating and cooling phases may be graphically determined
(Aiba, Humphrey, and Millis, 1973). Furthermore, in any real case, a deviation
from ideal plug flow characteristics can be considered with the help of the
dispersion model (Equ. 3.3); a reaction term representing the kinetics is added.
Using a dimensionless number (DaI , the Damkoehler number of first degree)
for the kinetics
and choosing the Danckwerts boundary conditions
dN =0 at z = 1 (6.90a)
dz
dN
- + Bo(1 - N) = 0 at z --+ 0 (6.90b)
dz
one gets a solution of Equ. 3.3b with a reaction term for Bo --+ high expressed
as
N(L) ( Dal) (6.91)
No = exp - Da. + Bo
This equation can generate a nomogram to evaluate tSt as a function of Bo at

various kd values (Aiba et al., 1973).
A problem similar to sterilization occurs in the treatment offood with heat
for purposes of preservation. However, the situation is more complex due to
the multiple components present and the requirement for high quality stan-
dards (cf. Fig. 5.79). The technical advantages of a CPFR must be considered
when new technology (such as, for example, the fluidized or spouted bed
reactor) is introduced (Baxerres, Haewsungcharern, and Gibert, 1977).
6.4.4 ApPLICATIONS OF CPFR

The CSTR cascade is often used as a substitute for CPFR behavior, and
tubular reactors themselves are rarely found in bioprocessing. Exceptions are
waste water treatment in oxidation ponds, river analysis (Metcalf and Eddy,
1972), sterilization technology, and fixed-bed reactors filled with immobilized
enzymes for bioconversions (Lilly and Dunnill, 1972).
Tubular reactors as horizontal equipment have been summarized by Green-
shields and Smith (1974) and by A. Moser (1983b, 1985b); applications vary
from fermentation processing with flocs, including waste water treatment, to
unconventional bioprocessing with immobilized cells, solid substrates, shear-
sensitive tissue cell cultivation, and phototrophic organisms (Pirt et al., 1983).
Figure 6.33 represents the advantages of a CPFR in case of waste water
treatment by showing the drastic reduction in volume (F. Moser, 1977).
6.4.5 ONE-PHASE (LIQUID) REACTORS WITH ARBITRARY

RESIDENCE TIME DISTRIBUTION AND MICROMIXING
6.4.5.1 Graphical Methods
Reusser (1961) described a graphical procedure for predicting the yield of a
nonautocatalytic process taking place in a continuous reactor in which the
residence time distribution (RTD) may be freely chosen and the kinetics of the
V FIGURE 6.33. Demonstration of the

% advantage of CPFR used in bio-
100
A logical waste water treatment by
80 showing the volumes needed in
B
aeration tank (A) and sedimenta-
60
tion tank (B). Two techniques are
compared: the conventional tank
40
B (STR) with air and a tube reactor
B
A B (CPFR) aerated with air or oxygen
20 - B
A RA and using the concept of reaeration
o A A ~A of the sludge (RA) (F. Moser, et aI.,
_ _ 0_ _ 1979).
-- s-m-o--"---CPFR
1- . P/Pend
1_ .
AREA YIELD
b
c
FIGURE 6.34. Calculation of the expected yield in a normal catalytic process: (a) Double
plot of concentration/time for a discontinuous process (P/Pend)/(t/t cnd ), and curve for
a continuous reactor with variable residence time distribution. (b) Evaluation of yield.
(Adapted from Reusser, 1961.)
equivalent discontinuous process are known. No mathematical formula is

necessary in this method.
In Fig. 6.34a, two curves are plotted that show the time course of the
discontinuous process kinetics and the residence time distribution in a con-
tinuous reactor of arbitrary RTD. Normalized values are used for ease of
comparison (Pend = final concentration at time tend; Co = total concentration
of the pulse used to initiate the RTD measurement).
This plot can be used to eliminate that t axis: For each value of t the values
from the kinetic and the RTD curve are read and are plotted against each
other in a separate diagram (Fig. 6.34b). The area under the curve gives directly
the yield to be expected in the steady-state condition of a continuous process.
This graphical method works only in cases where the reaction rate decreases
FIGURE 6.35. Predicted produc- rp

tivities of different novobiocin
fermentation systems expressed
in terms of means residence time
N
of the overall process (cell growth
CD
in tx and production in t p ) with 9 NCSTR
2
variation of the number N in 1
a cascade (NCSTR) compared
with the productivity in a batch
process (DCSTR). (From Reus-
ser, 1961.)
with a decrease in the concentration (cf. Fig. 6.31a), and it is used in situations
such as, for example, predicting the yield of an antibiotic-producing process.
Predictions based on this concept are shown in Fig. 6.35 (Reusser, 1961).
For autocatalytic and biotechnological processes (Fig. 6.31 b and c), a
basically similar method can lead to the same goal; the solution, however, can
only be obtained using mathematical formulas (cf. Equs. 6.92 and 6.94).
6.4.5.2 Calculation Methods

The simplest concept of a reactor is that each independent element of fluid
volume behaves as a DCSTR. This is the case of micromixing with total
segregation, J = 1 (cf. Fig. 3.1b). The fluid stream leaving the reactor is then
an average of all the individual DCSTR elements, each of which is present for
a different length oftime in the entire system. In mathematical form, the output
stream concentration is given by Equ. 6.92 (Danckwerts, 1958):
Cts = I'' cdc(t) f(t)dt (6.92)
where f(t) is the residence time distribution measured with the pulse method
(cf. Equs. 3.5 and 3.10) and Cdc is the concentration of the component in a
DCSTR.
Using sterilization as an example (cf. Equ. 5.275), the concentration value
for the DCSTR is expressed
(6.93a)
so that the sterilization process in, for example, a CSTR with a time distribu-
tion as in Equ. 3.11 may be calculated as
(6.93b)
When the second extreme case of micro mixing (from Fig. 3.1) is considered
-maximal mixing (mm) with J = O-the liquid volume elements in the reac-
tor differ from one another in their "life expectancy" ()" = 1 - t). From
a differential conservation of mass equation (and simplification), one has
(Zwietering, 1959)
dc f(t)
dt = r + 1 _ F(t) . (c - co) (6.94)
For a first-order reaction, Equ. 6.94 is reduced to the form of Equ. 6.92.
These equations for total segregation and maximal mixing have been used
for predicting microbial growth with Equs. 5.38 and 2.14a in a CSTR and in
a CPFR (Fan et al., 1970, 1971; Tsai et al., 1969, 1971). The result is presented
in Fig. 6.36; the plot is similar to Fig. 6.1 b with Xo > O. In the region 0 < I1-max'
t < 2, the concentration of x in a CSTRmm is higher than the corresponding
concentration in a CSTR ts or in a CPFR. The influence of micro mixing is clear
here. Further, in agreement with Fig. 6.28, one sees that the utilization of
substrate S beings earlier in a CPFR than in a CSTR and that somewhat
higher X concentrations are present in a CPFR at a given value of Ks.
The question of micro mixing is meaningless when ns = 0; similarly, micro-
mixing is meaningless when Its = 1. Model studies of micromixing are
therefore always undertaken using second-order chemical reactions (n = 2)
(Danckwerts, 1958).
,,/--- --- - XpF
/" XST.mm
/
/
I
/
_---
-------- --- XST, t 5
'~~
--------SST.ts
~,~
'-,
L______ - L_ _ _ _ _ _ ~======~~====~==SS~T~.mm
. .-- flmax-t
2 3
FIGURE 6.36. Plot of the dimensionless concentration of cell mass x and substrate s for
a continuous culture as a function of the dimensionless mean residence time t as in
Fig. 6.1 b with Xo > 0: Calculated comparison between a CSTR with maximum mixing
(ST mm) or one with total segregation (ST,sl and a continuous plug flow reactor (PF),
assuming Monod kinetics with a death rate kd (Tsai et aI., 1969).
6.5 Recycle Reactor Operation 351
6.5 Recycle Reactor Operation

In considering continuous operation, recycle loops are often included (Her-
bert,1964), with improved performance, that is, higher treatment rate, smaller
vessel size, higher conversion. As a means for resource conservation and
recovery as well as for environmental problems, recycle systems have become
increasingly important.
6.5.1 PERFORMANCE EQUATIONS OF RECYCLE REACTORS

6.5.1.1 The CSTR with Recycling of Cell Mass
Recycling cells from a CSTR effluent provides a means to continually inocu-
late the vessel and to add stability to the reactor, minimizing the effect of
process perturbation. The productivity of a CSTR may be increased remark-
ably by recycling cells when the cells to be recycled are first concentrated by
the factor 13 = Xr/X in a sedimentation unit. With a recycling stream F" and
with r = FrIF, the conservation of mass equations are
Change = inflow + recycle - outflow + growth
and for x and s
dx
dt = D Xo + r' D Xr - D xex(1 + r) + f..l' x (6.95)
ds f..l'X
dt = D'so + r' Ds - D's ex (1 + r) - y (6.96)
Hence in the steady state

f..l
s=K s- - - (6.97)
f..lmax - Jl
y
x = -(so - s) D (6.98)
Jl
and
Xex = Y(so - s) (6.99)
In contrast to the CSTR with no recycling, where Jl = D, the CSTR with
cell recycling (rCSTR) realizes
f..l = D(l +r - r' 13) (6.100)
and in this case it is possible to operate with higher concentrations of cells
XrCSTR = xcsTR(1 +r - r' (3) (6.101)
A schematic diagram of such a recycling system is shown in Fig. 6.37a. The
equations are plotted in Fig. 6.37b together with the output of cells D' xex; for
!-------------___ x
B= ~
X
a
b
~ _ _ _ _~ _ _ _ _ ~_ _ _ _ _ _ ~_ _ _ _ _ L _ D
FIGURE 6.37(a). Flowchart for a continuous stirred vessel with recycling of cell mass
(recycling stream Fr ), with the cells concentrated in a settling vessel (concentration
ratio, f3). (b) Theoretical relation between cell mass concentrations, X, and flow rate,
D. For comparison, the value for a one-stage CSTR are shown (---).
o
FIGURE 6.38. Effect of recycling biomass concentration Xr on a CSTR in a diagram of
effluent substrate concentration Sex versus mean residence time "f(r = recycling ratio).
(Adapted from Andrews, 1972.)
comparison, curves for cell mass concentration and output of cells in a single
CSTR with no recycling are also plotted (dotted lines). As can be seen, a
dilution rate greater than maximum growth rate may be employed, increasing
overall productivity. The critical residence time is
1 l+r-r'/3
terit = D. = (6.102)
ent flmax
An estimate of the effect of cell mass concentration and recycling on process

economics may be obtained by plotting effluent substrate concentration for
several different values of x" as shown in Fig. 6.38. Increasing Xr has the
effect of minimizing Sex and also of making Sex less sensitive or more stable
with respect to variations in the flow rate of constant-volume reactors with
residence times considerably below the washout value (Andrews, 1971).
6.5.1.2 Cascade of Reactors with Cell Recycling,

For the case of cascade ofreactors with cell recycling, Powell and Lowe (1964)
derived a performance equation that is analytically intractable. Nevertheless,
a simple formula of the critical dilution rate DedI can be calculated in the case
of a cascade of CSTRs as
DedI = [(1 _~ ~~/N)N ] J1max Ks s~ So (6.103)
illustrating how the critical dilution rate depends on the degree of feedback r,
the number of stages N, and the concentration of limiting substrate.
Toda and Dunn (1982) emphasized that a combination of a backmix and
a plug flow fermenter with recycle streams provides better performance than
does a single CSTR for the continuous production of a substance that depends
on the maturity of growing cells (see Equ. 5.136). A typical example is shown
in Fig. 6.39, and it illustrates that the productivity of cell mass (Fig. 6.39a), of
D.i
,-,
I " ,,/ \
,
1
a b
&U..Io........~_D
D.p
," \\
,
I
\
\
\
\ ,
0
C
D'P
L '\\ 0 d
FIGURE 6.39. Productivity offermentation product in combined systems ofCSTR and

CPFR (cf. Fig. 3.lOd) in comparison with a single CSTR (----) in four different cases:
(a) cell mass production following Monod kinetics, (b) Luedeking-Piret-type product
formation, (c) productivity of repressible products, and (d) maturation time-dependent
products. (Toda and Dunn, 1982.)
products formed according to Luedeking-Piret logistic kinetics (b) (Equ.

5.122) and of repressible substances produced (c) according to the kinetic
model of van Dedem and Moo-Young (1973) is still greater in a CSTR. These
facts represent additions to the concept shown in Fig. 6.32.
6.5.1.3 Plug Flow Reactors with Recycling

CPFRs containing microbial flocs can only be operated continuously on a
recycling basis or with a CSTR as an inoculum reactor; otherwise washout
will occur. For a CPFR with recycling of nonconcentrated cell mass, a perfor-
mance expression was given by Levenspiel (1979). For Xo = 0
.- _ (
Ilmaxt-r+ 1) (Kg In So
So
+
r' S
r' S I 1+
+n
r
r) (6.104)
When material is to be processed to some fixed conversion ( in a CPFR with

recycling, a particular optimum recycle ratio r should exist. At this point
reactor volume is minimized. The final solution is yielded by putting ax/or = 0
in Equ. 6.104:
Kg In So + r' S + In 1 + r = 1 + r Ks +_ (6.105)
So r .S r r So + r' S r
This equation can be solved by trial and error. The optimum r is found to be
a function of Kg/so and s/so. A graphical procedure for the determination of
the optimum r is shown in Fig. 6.40 (Levenspiel, 1979). One has to try different
values of (i.1 until the two dotted areas are equal. This means that the r- 1 at
the feed entering the reactor is equal to the average r- 1 in the reactor, or
mathematically
(6.106)
-1/r. ~L.C_i,1_'-_-_-_-_-=-~-_-_-_..J_...J1
f -i,in[i,l' Fr . ci7ex F
1
~ave,R FIGURE 6.40. Optimization of conver-

sion in a recycle reactor in a plot of
:
r'l1 c i ,1/ reciprocal rate l/r; versus conversion
(; (cf. Fig. 6.32) by setting the shaded
,.
t
~. areas equal to each other according to
~i,in ~':1I, \ex 1 Equ. 6.106 (Levenspiel, 1979).
The value of r corresponds to the concentration according to (cf. Fig. 6.40)

Cj 0 - r' Cj , ex
C -'
j,l - 1+ r (6.107)
At too high a recycle ratio, the average rate is higher than the rate in the feed
entering the reactor, and vice versa. In Fig. 6.41 the optimum value ropt is
plotted as a function of Ks/so and s/so. For a CPFR with recycling of
concentrated cell mass, Grieves et al. (1964) derived a general equation that
can be used for computer simulation of the basic behavior:
- fJmax'
-
t = (1 + r)
{ (1 + r)Ks
[r' P(so - s)/1 + r - r' PJ + So + r' s
x In s'r' P - I n1 -+-r} (6.108)
So + s r r' P
As the critical residence time is approached, s increases much more sharply
Dcrit r---r-r-----.-----...,
A2
X:a;;..,..=o-rt-t
L - I_ _ --'i I 5
X
2 3 4 _ r
~------------~r~
6.42
Dcrit.CSTR . - - - - - - - - - - - - . . . ,
10
100 Dcrit.cPFR
1
o~-~--o~.2--~--H-S/SO
6.41 6.43
FIGURE 6.41. Graphical representation of the trial-and-error solution of Equ. 6.105

showing the dependence of optimum recycling ratio ropt in a CPFR with recycling on
substrate concentration s/so as a function of varied Ks value. (Adapted from Levenspiei,
1979.)
FIGURE 6.42. Effect of biomass concentration on the washout of a CPFR shown in a
graph of critical dilution rate Decil in dependence on recycling ratio r at varied values
of concentration factor f3 (Atkinson, 1974, with permission of Pion, London.)
FIGURE 6.43. Comparison of the washout flows of a CSTR and CPFR by showing the
ratio of critical dilution rate Deril depending on recycling ratio r with variation of the
concentration factor f3 (Atkinson, 1974, with permission of Pion, London).
and Xex decreases much more sharply than is the case for the CSTR (Fig. 6.26).
Equation 6.108 indicates the general improvement in performance with
increased recycling stream. The relationship between washout flow and re-
circulation can be deduced from Equ. 6.108 be setting sols equal to unity
(corresponding to zero conversion). According to Atkinson (1974)
D = _1_ = Jlmax s_
__
(6.109)
c t.,rit (1 +r)ln[(1 +r)/f3r] Ks+s
The improvement in performance when 13 is increased is shown in Fig. 6.42,
where both factors of the recycling ratio r, and 13, the cell concentration factor,
influence the performance of CPFR. A comparison of washout flow can be
obtained from Equs. 6.102 and 6.109 (Atkinson, 1974):
l+r .ln1+r (6.110)
1 +r-rf3 f3r
whiCh is depicted in Fig. 6.43.
6.5.2 ApPLICATIONS OF eRR

In fermentations and other bioprocesses, process stream recycling is inti-
mately concerned with enhanced conversion. Such recycling schemes usually
seek to reduce operating costs by increasing capital investment. The three
major possibilities where recycling loops can be expected in the fermentation
industry involve (a) processes employing gaseous substrates, including pure
oxygen; (b) processes employing spent medium component and/or process
water reutilization; and (c) processes designed to improve fermenter produc-
tivity by using elevated cell concentrations (Hamer, 1982).
In biological waste water treatment, the recycling of sludge has long been
used (Metcalfand Eddy, 1979). As a consequence of recycling, reactor stability
and substrate conversion are increased. Other important cases of the applica-
tion of recycling operation are known from process design optimization for
continuous ethanol production (e.g., Cysewski and Wilke, 1978). Generally,
various systems for cell recycling are available: internal filtration, internal
sedimentation, monostream, and external recirculation (Pirt, 1975). Recently
a computer simulation study of ethanol fermentation with cell recycling
showed several interesting facts (Lee et ai., 1983). The cell growth equation
with an inhibition term (cf. Sect. 5.3.5) predicts that the relative productivity
of the recycle reactor will increase drastically (400-500%) as the bleed stream/
feedstream ratio B/F decreases until the cell concentration in the reactor
reaches the maximum value x max .
Similar optimization studies were experimentally verified for ethanol pro-
duction with Zymomonas mobilis (e.g., Charley et ai., 1983). A two-stage
CSTR system with cell recycling in the second stage resulted in high final
alcohol concentrations (> 100 gil) at quite high overall volumetric produc-
6.6 Gas/Liquid (Two-Phase) Reactor Models in Bioprocessing 357
tivity. Moreno and Goma (1979) used a cascade of eight CSTRs in a series
and a yeast strain; with yeast the increase in cell concentration by recycling
is not accompanied by a corresponding increase in ethanol productivity due
to irreversible inhibitory effects on the recycled cells. Similar trends were
observed with Z. mobilis by another group (Boks and Eybergen, 1981), where
biomass production decreased while ethanol productivity did not change
markedly.
Last, but not least, it should be mentioned that some other newly designed
bioreactors are based on the concept of a CRR. A tubular loop reactor with
a G-phase separator chamber or a cyclone was installed successfully in bench
and pilot scale (Russell et a!., 1974; Ziegler et a!., 1977, 1980). Another loop
reactor that also realizes plug flow in the L phase to some extent is the
horizontal circular ring reactor (Herzog et al., 1983; Laederach, 1978). The
cycle tube cyclo~e reactor is a vertical construction of a loop reactor that is
a highly efficient aeration device with low power consumption (Liepe et aI.,
1978; Ringpfeil, 1980). These configurations are compared in a recent review,
together with the fields of application (A. Moser, 1985b).
Finally, several other contributions to the literature suggest that the princi-
ple of recycle operation is being acknowledged in bioprocess research and
development (Adler and Fiechter, 1983; Blenke, 1979; Bull and Young, 1981;
Constantinides et aI., 1981; Lippert et aI., 1983; Seipenbusch and Blenke, 1980;
Stieber and Gerhardt, 1981a,b).
6.6 Gas/Liquid (Two-Phase) Reactor Models in

Bioprocessing
The formulation of a gasjliquid reactor model may be done by considering
the analogous situation in chemical process engineering (cf. Sect. 4.5.4, espe-
cially Fig. 4.46). The importance of OTR enhancement in analogy to chemical
processing (e.g., Nagel et aI., 1977, 1978) can be stressed by showing Fig. 6.44.
For a first-order reaction, the relative conversion 'rei is a function of both
factors, the enhancement 1fTR (Ha, cf. Equs. 4.115 and 4.112) and the ratio of
hinterland HI (cf. Equ. 3.59). In the case of a slow reaction ('7TR = 1), bio-
reactors with a high hinterland will give satisfactory conversion (e.g., bubble
columns). Stirred tanks are more flexible (10 2 < HI < 103 ) but do not have
enough interfacial area to achieve high conversion in the case of fast reactions.
Only when using bioreactors with low HI values (high a values), for example,
jet nozzle reactors and thin-layer reactors, can sufficient conversion be reached
with "fast" reactions ("ITR > 1; Ha > 0.3). Thus, the choice of optimum bio-
reactor can only be made by a complete analysis and full understanding of
this type of interaction between physics (transports in bioreactor) and biology
(kinetics of metabolic reaction).
It was shown with the aid of modeling that theoretically only in aeration
1
~rel
103 L -____~~~_ __L~_ _ _ __L~_ _~
1 101 102 103 104

-HI
FIGURE 6.44. Dependence of the relative conversion 'rei on the degree of hinterland HI
in a GIL reactor (see Equ. 3.59) [bubble column, Be; stirred vessel, ST; thin-layer
reactor, ThL; injector (jet nozzle) reactor, IN] for an aerobic bioprocess with various
fast reactions, as represented by the oxygen transport enhancement factor, E (see Equ.
4.115). (From Nagel et aI., 1972.)
systems with low kLl values at high x and high qO,max in the case of no S
limitation and low ms in the region of no = 1 could bioenhancement contri-
bute significantly to OTR (A. Moser, 1980b). The results of this modeling are
in agreement with trends reported in the literature and are supported by recent
investigations using the absorption system carbon-dioxide-phosphate buffer-
carbonic anhydrase for the measurement of kLl and a, which indicate that
absorption enhancement is completely in line with ordinary theory (Alper et
aI., 1980b).
It must be stated here that it is hard to verify this effect in a reliable
experimental method due to the difficulty of operating with microbial cells at
fixed kinetics and of measuring kLl and a separately (cf. Fig. 3.43) Kung and
Moser (1988).
6.7 Biofilm Reactor Operation

Due to the adhesion capacity of microbes, biofilm formation must be taken
into account in river analysis as well as in the chemos tat or fixed beds
(percolating or trickling filter). The significance of microbial films in fer-
menters has been extensively studied and reviewed (Atkinson, 1973, 1974;
Atkinson and Fowler, 1974; Atkinson and Knights, 1975; Charaklis, 1981;
Harremoes,1978).
6.7 Biofilm Reactor Operation 359
6.7.1 POTENTIALITIES OF BIOFILM REACTORS

Process engineering possibilities of biofilms have been compared with those
of floc processing in Table 3.1. As a consequence of these properties, process
development using cell support systems seems to be promising (Atkinson and
Kossen, 1978; Atkinson et aI., 1980). The potential advantages are as follows
(Moser, 1985a):
- As washout is not possible, high throughputs can be realized in reactors.
Biomass holdup is independent of throughput.
- Due to the adhesion of cells on support surfaces, high biomass concentra-
tions can be realized.
- Due to the adhesion to solid particles, the principle of fluidization can be
introduced to bioreactor operation. This leads to the advantage of better
transport processes, especially at LIS interface (turbulence) and in the S
phase (abrasion or scouring).
- Sometimes the appearance of transport limitations or concentration gra-
dients may be advantageous, as in these cases:
a. With mixed populations-different species may occur at various depths
within the particle, for example, nitrifiers near the surface and denitrifiers
toward the center (Eggers and Terlouw, 1979).
b. When using cells that must be "mature" before the product is produced
(cf. Equ. 5. 136)-retention ofthem within a particle may be beneficial, for
example, citric acid and secondary metabolite production.
c. When using S-inhibiting media, the most favorable S concentration may
occur toward the center of the film.
- A special advantage ofbiofilm reactor operation in practice comes from the
fact that the performance is independent of film thickness when thick films
are used. However, as in thick biofilms generally Sand O 2 limitations occur,
which result in a loss of viability of the cells. Thereby, the adhesive bond to
the support surface is weakened so that sloughing will occur, which will
disturb the reactor performance.
- As a consequence of the foregoing problem, the control ofbiofilm thickness
is of central importance. The thickness can be controlled by means of
mechanical scraping or abrasion by friction as in the self-regulation effect
at high hydrodynamic stress (cf. Sect. 3.6).
- A unique benefit of biofilm reactors for research purposes is the fact that
due to the distinct separation between the L and the S phase (difference in
density Ps - prJ, high relative velocities can be realized. As a result, external
transport limitation can be excluded or easily studied simultaneously with
internal transport limitations in the case of uniform and controlled biofilm
thickness. In this respect, biofilm reactors are superior to the conventional
STR.
Last, but not least, pellet processing should be mentioned here (Metz and
Kossen, 1977; cf. Sect. 5.8.2). Usually investigators consider pellet formation
as an undesirable feature, leading to inhomogeneous mycelium. Nevertheless,

controlled pellet formation offers the advantage that pellet suspensions have
a much lower viscosity than a filamentous broth (pulp). This leads to pellet
processes that are energetically more efficient than a traditional pulp-like
process. This was shown with a thermodynamic approach to calculation of
the thermodynamic efficiency (see Equ. 2.29). The activity loss in the center of
a pellet is very well compensated for on a macroscopic level by the reduced
power requirement for such a process (Roels and van Suijdam, 1980).
6.7.2 PERFORMANCE EQUATIONS OF BIOFILM REACTORS

Complicated models for film bioreactors may be simplified when pseudo-
homogeneity can be assumed (see Sect. 4.2)-that is, when the S-concentration
profile is flat throughout the cross section of the trickling filter.
6.7.2.1 Pseudohomogeneous Reactor Modeling of Fixed Bed Reactors

6.7.2.1.1 Kinetics as the Rate-Determining Step
There is a concentration gradient present in the fluid flowing over the length
of the reactor at a rate F [m 3/h]' This requires dealing with the mass conser-
vation for S using differential volume elements, as in Fig. 3.38 and Equ. 3.95
(with no recycling).
F(s + ds) = Fs - rsdVx (6.111)
or
fJ.
Fds = --'x'dVx (6.112)
Y
The important value for X is not the total mass of X; rather it is the active
mass of X in the external surface of the film. The volume of the active biomass,
Vx , may be calculated as the product of several factors: the differential thick-
ness, dz; the cross-sectional area, A; the specific surface area ofthe falling body,
as [m2/m3]; and the film thickness, d
dVx = as'd' A 'dz (6.113)
The final equation for S use as a function of depth may be written
- ds =!!:. a .X .d . A .! (6.114)
dz Y s F
Using Monod kinetics, the mass conservation becomes
F . ds + fJ.maxxsasdAd z = 0 (6.115)
Y(Ks + s)
With constant film thickness and constant K s , the integration with bounds
z = 0 at So and z = z at Sex gives Equ. 6.116:
--r--+---"'7"~---------~N-SEX
FIGURE p.45. Graphical method of parameter estimation in case of Equ. 6.116

representing a pseudo homogeneous approach to biofilm processing, Kornegay and
Andrews (1969) and Kornegay and Andrews (1968).
InSex = sin - Sex _ (JLmaxdx,x)ax.A'Z (6.116)

Sin Ks Y F Ks
This solution describes the S consumption in the fixed bed with essentially
plug flow behavior and shows that a plot ofln(sex/sin) versus Sin - Sex should
give a straight line with a slope of 1/Ks and an intercept as indicated in Fig.
6.45 in agreement with Equ. 6.116. Thus, the parameters Ks and the term inside
the parentheses of the second term on the right side of Equ. 6.116 can be
evaluated from bench- or pilot-scale data in which Sin is varied at a fixed flow
rate. Both values are pseudokinetic parameters, which depend upon liquid
flow rate. Nevertheless, Equ. 6.116 can be used to successfully predict effluent
conditions for various values of F, Sin' and total length of the fixed bed.
Modifications ofthis solution are needed for the case where dxis not constant,
which will occur at substrate concentrations below about 300 mgjl. Here dx
will increase linearly with S(Harris and Hansford, 1976). Thus, instead of Equ.
6.116, the S balance must be modified, leading to a somewhat different solution
given by
(6.117)
where d~ is a proportionality constant (d x = d~ . s).

6.7.2.1.2 Transport Factors as the Rate-Limiting Step
A differential conservation of mass equation (Equ. 6.118) is found in the case
where S utilization is determined by the rate of substrate transport through
the liquid-solid interface (kL2 in Fig. 4.28). The latter is the more likely case
with trickling filters
F ds = - kL2 . as(sL - st) A . dz (6.118)
With st ~ 0, integration of Equ. 6.118 leads to
Sex k L2 az
In-= - - - - (6.119)
So FjA
According to La Motta (1976)
kL2 = k~2 'vN (6.120)
with VS,L the surface flow velocity of the liquid phase, which is equal to (F j A),
so that
(6.121)
This result, which is based on a reactor model that assumes a mass transport
limitation, is substantiated by empirically based model equations that are
formally first order (cf. Equ. 5.159). These have the same form as Equ. 6.121,
in which (k~2' as), the apparent value kapp , and the power of VSL appear as
experimentally determined coefficients.
It is interesting to note that many empirical models have been developed
for fixed bed bioreactors, and that these models agree with this equation (e.g.,
Eckenfelder, 1966; Oleszkiewics, 1976, 1977). A practical design equation for
trickling filter processes was developed by Kong and Yang (1979) by combin-
ing the approach of Kornegay and Andrews with the practical aspect of sludge
age (sludge retention time) suggested by Kincannon and Sherrard (1974) and
applied by Bentley and Kincannon (1976). Sludge age eX' which can easily be
determined by any of the two independent process variables of the hydraulic
loading rate Bv and S or by the organic loading rate Bx and S, can be employed
to give
_Si_"_-_se_x F Y _ k (6.122)
x V d
This equation can be plotted to give the values of Y and kd from the slope and
the intercept. All performance characteristics of trickling filters were shown
to be a function of ex (e.g., removal, production, and volume index of sludge).
A mathematical model for percent removal of a pure, nonadsorbable,
biodegradable substrate in a submerged biological filter was developed by
Jennings et al. (1976), again using a Monod-type relation for biokinetics.
Problems in the interpretation of removal kinetics are encountered in this case
(Ottengraf, 1977).
Another pseudo homogeneous model of biofilm reactor operation was
derived by Atkinson and Davies (1972) on similar assumptions to quantify
the behavior of the CMMFF (cf. Fig. 3.42). For this case of a completely mixed
reactor system, the balances on X and S, followed by algebraic combination

and rearrangement, lead to the performance characteristics of fermenters
containing immobilized cells:
(6.123)
with A and B being coefficients defined, in the case where no thick film is
present, as
A = sin
Ks
(1 _ Ilmax'
F
V) (6. 124a)
and
B = 1_
Ks
sin + Ilmax (Sin
F KS
+~)
Y'K s
(6. 124b)
The solution of Equ. 6.123 takes the form
~=
Sin
f( F
V' Ilmax '
Sin
Ks '
y~iKn s) (6.125)
and means that the conversion efficiency depends upon a dimensionless flow
rate, inlet concentration, and biomass holdup. The algebra of Equ. 6.125 has
been extended to include both internal and external transport limitation
(Atkinson and Mavituna, 1983). The setup of true heterogeneous models for
this situation of interacting biokinetics and external and internal transport
was broadly discussed in Sects. 4.5 and 5.8.
The final equations for the solution of this interacting system are sometimes
given in the form of the pseudohomogeneous rate Equ. 6.121, substituting for
the factor (k~2 . aLls) the rate constant for biological reaction kr multiplied
with the effectiveness factor 11r (cf. Equ. 4.51)
Sex 11. k r ' Z
-=exp---- (6.126)
Sin VSL
The effectiveness factor 1'Ir varies only with the biofilm thickness and the
support particle size at constant voidage of the fluidized bed. This leads to the
conclusion that an optimal biofilm thickness exists for fluidized bed biofilm
reactors (Kargi and Park, 1982). The effect ofbiofilm thickness on the perfor-
mance of a fluidized bed reactor was also examined by Shieh (1981), who
showed the same results. Neither maintenance of thicknesses with 10 to 100
Ilm nor the highest biomass will be beneficial to optimum reactor operation.
Techniques for estimation of biofilm thickness were recently summarized
by Charaklis et al. (1982). An ideal film thickness equal to the penetration
depth of the limiting substrate or oxygen was stressed by Howell and Atkinson
(1976). A capillary microelectrode technique has been developed by Bungay
et al. (1969) for this purpose. The effect of growth rate on the biofilm buildup
has been observed by Molin et al. (1982). All of these developments may help
to elucidate problems of biofilm reactor operation.
The pseudohomogeneous biofilm model, given in Equs. 6.116 and 6.121,
can be readily adapted to quantify the behavior of a rotating biological disk
reactor. The setup of a similar balance equation for S using Monod-type
kinetics yields for the case of rds kinetics
Sex 1
(6.127)
Sin 1 + (J1.max dx ' x ax' t )/[Y(Ks + sex)]
Ifrds = LIS transport, then, similarly to Equ. 6.121, the solution of the balance
equation is
Sex 1
(6.128)
Sin 1 + kL2 . aLls . t
for each stage, where t = V/F, the space time.
Wu et al. (1980) analyzed and modeled a biodisk system with six stages, and
showed that the conversion efficiency is directly associated with t and is
inversely affected by Sin' temperature T, and stage number N:
S
~ = 14.2' EZ' sS Sg68To.24 (6.129)
Sin exp(0.32N)
6.7.2.2 Unified Performance (Heterogeneous) Model of Fixed Bed

Biofilm Reactor
Pseudo homogeneous models ofbiofilm reactors will fail in cases where deeper
understanding and interpretation are needed. There are three basic require-
ments for a heterogeneous model for biofilm reactors (Rittmann, 1982):
1. Quantification of rs, S utilization rate, in the case of interactions between
reaction and internal or external transports. This problem of solving Equs.
4.76 and 4.101 simultaneously was discussed in Sect. 4.5. These solutions
give steady-state S utilization but do not consider growth or decay of the
biofilm. Adequate modeling of biofilm reactor operation must incorporate
this fact.
2. Quantification ofrx, the growth. A specific decay coefficient kd (cf. Equ. 5.76)
has been shown to give experimentally valid results with fluidized beds
(Andrews and Tien, 1982). Rittmann and McCarty (1980) have defined a
steady-state biofilm as one having no net growth or decay for the entire
biofilm depth. This dynamic constant thickness can be calculated by equat-
ing growth due to substrate flux n~ with the maintenance decay of the entire
biofilm. Thus
dr = n~' Yxls (6.130)
k d Xc
The significance ofthis prediction was questioned by Arcuri and Donaldson
(1981). Supplementary to this effect of maintenance decay, biomass is also

lost from the biofilm through the action of shearing, in which the stress
of water flowing past the biofilm pulls away part of the film. This effect
can formally be quantified with an apparent decay coefficient k t [h-1J
(Rittmann, 1982) to be incorporated into Equ. 6.130 if shear losses become
significant as, for example, in a fluidized bed biofilm reactor (FBBR). A key
concept of the steady-state solution is the existence of a threshold concen-
tration Smin' below which no significant steady-state biofilm activity occurs
(see active depth <>cril' Equ. 4.89). At bulk concentrations greater than Smin'
n~ and df are determined by simultaneously solving Equs. 4.77, 4.101, and
6.130 (Rittmann and McCarty, 1980). At high enough S, the steady-state
flux becomes equal to the flux into a deep biofilm, where the reaction order
0.5 < ns < 1 (cf. Fig. 5.73).
3. Formulation of a proper reactor model and incorporation of kinetic expres-
sions for rs and rx. Using the fundamental law of conservation of mass (cf.
Equs. 4.76 and 4.101), the general partial differential equation of S concen-
tration in a control volume is
dV'GL -
os = os
-vz-dV + Defr -
02S
dV -
,
rsdV + arnsdV) (6.131)
ot OZ OZ
2 (GL
where GL is the bed voidage or porosity of the medium and the terms inside
the parentheses represent removal reactions brought about by biofilm with
ar' n~ and suspended cells rs. Equation 6.131 can only be solved numerically,
and numerous techniques (e.g., Remson et al., 1971) are available.
The development of such unified models was presented by Rittmann (1982)
for different biofilm reactor types, that is, for a completely mixed reactor,
once-through fixed bed reactor, and once-through FBBR without and with a
recyclestage. Although Equ. 6.131 has been solved numerically for all reactor
types, each reactor requires an appropriate value of Sin, a submodel to calcu-
late n~ and dr, and an appropriate number of reactor segments (e.g., N = 1
for one CSTR and N = 12 for FBBR). These terms differentiate among the
reactor models and are discussed in Rittmann's original paper.
An important consequence of this unified model approach is that it can be
applied to all reactor configurations, since it is based on common principles
for the dynamic interaction of kinetics and transports. These results demon-
strate that simple loading criteria cannot predict the performance of a variety
of biofilm operations because they ignore the interactions among rs, rx , and
reactor balance.
Another advantage of quantifying the dynamic interactions is that transient
conditions (e.g., conditions of active microbial growth) can be studied in
various biofilm reactors, as was demonstrated by Andrews and Tien (1982).
As this unsteady-state model requires the use of computer techniques, Andrews
(1982) presented a method by introducing a variable transformation that
eliminates the need for a computer solution. However, several limitations to
the modeling results are known, so caution must be exercised in applying the
unified model approach, that is, non-steady-state loading (fluctuations in Sin

and vz ), dual-substrate limitation (e.g., Reuss and Buchholz, 1979), the effect
of additional biomass on particle-settling velocity (which is different for, e.g.,
tower fermenter and fixed bed reactors), and some practical considerations
(bed expansion, clogging, wall effects, and shearing). The use of non-monosized
support particles is recommended as an alternative to a tapered tower fermen-
ter, as normally a fixed bed bioreactor is superior due to almost constant cell
concentrations throughout the bed (Andrews, 1982).
6.7.2.3 Fluidized Bed Biofilm Reactor (FBBR)

Recently, the application of fluidized beds to bioprocessing stimulated the
modeling of FBBRs. In this complex case, however, models of the hydraulics
must be combined with a model ofbioftlm kinetics, including mass transport-
affected substrate conversion. This truly heterogeneous case of reactor opera-
tion will be presented here.
The unified model approach, according to Rittmann (1982), predicts that a
once-through FBBR can achieve performance superior to mixed ~)TR and
fixed bed reactors because the bioftlm is evenly distributed throughout the
reactor while the liquid flow still exhibits plug flow behavior. This increases
the overall effectiveness of the FBBR, in which effiuent concentrations less
than Smin will be obtained.
Because the principal advantage of the FBBR is the reduction in reactor
size caused by the increased biomass concentration, an understanding of the
factors affecting x is essential. A theoretical value for the reactor biomass may
be determined by development of a hydraulic model on the basis of the
principles of LIS fluidization. Mulcahy and La Motta (1978) developed a
mathematical model for a FBBR. Prediction of S conversion requires again
the simultaneous solution of the reactor flow equation coupled with bioftlm
effectiveness equations. Solution of these equations, however, is possible only
after the parameters of bed porosity e and bioftlm thickness ~ have been
specified. An algorithm is proposed for this fluidization model that correlates
these through a parameter referred to as the expansion index, that is related
to the terminal settling velocity of the bioparticles.
The dispersed plug flow model was used to describe the flow behavior in
the FBBR (cf. Equ. 2.3c, 3.3a, or 3.97). At steady state, the conservation
equation, using dimensionless parameters as in Equ. 4.79, becomes for the L-
phase expanded bed:
(6.132)
where '7r is the overall effectiveness factor, defined as in Equ. 4.51 and ri is
the reaction rate on a per-unit expanded bed volume basis. ri is related to the
observed rate of reaction on a per-unit bioftlm volume basis rS,eff by the
following simple expression
Yx' rs.effl z = Hx' A . r;-Iz (6.133)

with Yx the biomass volume and Hx' A the expanded bed volume. Thus
r;- = 1]r' ex . rS. ideal (6.134)
where ex is the biomass holdup (Yx/Hx' A), which can be conveniently expressed
in terms of bed porosity e and the operating parameter, the media volume Vm ,
as
Vm
ex = l - e - - - (6.135)
Hx'A
Substituting Equ. 6.134 for r;- and Equ. 6.135 for ex into Equ. 6.132 results in
an adequate reactor flow equation to be solved simultaneously with the
biofilm effectiveness equation. The equation for 1]r represents the solution of
the previous problem of internal transport with simultaneous enzyme reac-
tion, described in Equ. 4.81. The effect of external mass transport, or the
rate of S conversion, is included through the boundary conditions at the
LIS interface. Equation 4.81 and the accompanying boundary conditions
expressed in dimensionless form are
1 '~[(x + dm )3 d(S/Sd] _,/,2 S/SL - 0

[x + (d m /2b)]2 dx 215 dx 'I' (Ks/sd + S/SL - (6.136)
with
d(s/~d = Bi
dx
(1 _~) SL
atx = 1 (6. 137a)
and
atx = 0 (6. 137b)
x
where = rib - dm /2b = dimensionless bioparticle radial coordinate
dm = support medium diameter
Bi = Biot number
(cf. Equ. 4.110). The overall effectiveness factor, comparing rS.eff to rS.ideal, can
then be obtained by using Equ. 6.136 with Equ. 6.137 after integration for the
calculation of rS.eff' The following expression, obtained in this way and written
f
in dimensionless form, is the biofilm effectiveness equation
1 + KS/SL 1 S/SL (A dm )2 A
1]r = (d m /2bf+ (d m /2b) + 1/3 0 (Ks/sd + (s/sd x + 215 . dx (6.138)
that is needed together with Equ. 6.132 for the quantification of the perfor-
mance of FBBRs.
To predict S conversion within an FBBR using the mathematical model
presented, a total of 13 system parameters and six empirical correlations must
Design Parameters
VJ F,qs,max,Ks,DSL,DSX,sLlz=o
0\
00
Fixed Parameters
Sh/E , Re , Sc
Correlations
'--- BIOFILM MODEL
HX' A, Vm ' Pm' dm Eqs.6-136,6-138
Design Parameters ~ EFFECTIVENESS
F ,VL ,P L FLUIDIZATION E,6
sd- SL(Z)
Fixed Parameters MODEL ~ llr
p/6, CDIRe , nS/Re t I.....-.
Correlations REACTOR MODEL
Eq. 6-132
r----
dm ' A
Design Parameters
F, qs, max ' Ks ' s Liz =0
Fi xed Parameters
SolE ,Re , Remf,Remf/Ga
Correlations
FIGURE 6.46. Block diagram of the fluidized bed biofilm reactor model" system dependent parameters-inflow concentration of substrate S,
according to Mulcahy and La Motta (1978), containing reactor flow sLlz = 0; inflow rate, F.
equation, biofilm effectiveness equations, and fluidization model with 13 Empirical correlations:
system (fixed and design) parameters and six empirical correlations.
1. Biofilm density-biofilm thickness correlation (Pxlo).
Design parameters: reactor parameters-horizontal area, A; expanded
2. Drag coefficient-Reynolds number correlation (CD/Re).
bed height, Hx-and support media parameters-media density, Pm;
3. Expansion index-terminal Reynolds number correlation,
media diameter, d m ; total volume of media, Vm . Fixed parameters: liquid
Equ.6.139.
phase parameters-diffusivity of substrate S in liquid, DSL ; liquid density,
4. External mass transfer coefficient correlation, Equ. 4.1.
PL; liquid viscosity, VL -biofilm parameters-diffusivity of substrate in
5. Axial dispersion coefficient correlation (Bole resp. Re).
biofilm, Dsx; maximum rate constant, qs,max; Michaelis constant, Ks-and
6. Minimum fluidization Reynolds number correlation (Re_,/Gal.
be specified. The required parameters and correlations are presented in the

legend of Fig. 6.46, where a block diagram of the FBBR model is presented
(Mulcahy and La Motta, 1978).
The fluidization model indicated in Fig. 6.46 is not elaborated on here in
detail. An algorithm is proposed by these authors, as already stated, correlat-
ing bed porosity e determined through the expansion index nB' which corre-
sponds to the bioparticle terminal Re number ReI as a measure ofthe terminal
settling velocity VI calculated from Newton's law in the following manner:
With
nB = 10.35 Re l - O. 18 (6.139)
the resultant equilibrium bed porosity e is
e= (~ylnB (6.140)
The most common approach, however, in contrast to this strategy for the
setup of a fluidization model, is first to define an empirical correlation for an
isolated particle and then to extend it to cover multiparticle systems through
inclusion of a correction factor dependent on e, as demonstrated by Shieh et
al. (1981). For an isolated spherical particle system with defined characteristics
(d p , Pp ' PL' and v), these workers gave the following expression
e4 . 7 Ga = 18 Re + 2.7' Re1.687 (6.141)
Developing an equation for the biomass concentration, again starting with
an idealized bioparticle, Shieh et al. (1981) showed that
(6.142)
where Px is the dry density of biofilm and dp is the diameter of bioparticle.

Examination of Equs. 6.141 and 6.142 reveals that the biomass concentra-
tion in the FBBR is mainly a function of the bioparticle diameter (or biofilm
thickness) and of the hydraulic characteristics of the reactor that are under
the control of the design engineer: that is, expanded bed height, dm ; reactor
area perpendicular to flow, ~, which directly depends on V LS ' media volume;
and expanded bed height. This proposed model was shown to be capable of
predicting biomass concentration in the oxitron system FBBR. Support media
size and biofilm thickness are two important variables affecting the biomass
concentration. Shieh (1981) also developed a kinetic model for design purposes
of the FBBR, assuming zero-order kinetics and complete S utilization in the
outer shell of the biofilm, which may not be generally true. Integration of
similar concepts as previously described (Equ. 6.43 with rs = ko and a biofilm
effectiveness equation) yields in this case
8~55 = -kerr.I + 8g 55 (6.143)
where I is the mean residence time and
= 1 657(1 - e)(p' k )0.55. Do. 45

k err err I( d
~
2
)0.9 (6.144)
Thus, the S concentration profile through the reactor can be described by a

0.55-order rate equation (cf. Equ. 5.254), and this result was successfully tested
with data reported from Mulcahy and La Motta (1978) at different biofilm
thicknesses. Maximum S conversion was reached with an optimal support
particle size dp,oPI' However, it is important to note that kerr, a pseudo-
homogeneous rate coefficient, is actually a parameter rather than a true rate
constant, as it depends on the characteristics of both biofilm and substrate.
In contrast to the result from this work, Kargi and Park (1982) proposed
a theoretical analysis that indicates no dp,oPI value but rather an optimal
biofilm thickness for a given support particle size, that is, an optimal ratio of
biofilm thickness/support particle size (b/Rm) when eL is constant.
Using a Thiele modulus rP~ according to Equ. 4.111 (cf. Fig. 4.45), the
solution in this case has the form of the following equation (cf. Equ. 6.121)
valid for FBBR design
(6.145)
with H being the height of the bed and V SL being the superficial velocity of the
fluid [cm' S-1].
6.7.3 ApPLICATION OF BIOFILM REACTORS

While types of biofilm model reactors have been summarized in Fig. 3.42,
industrial-scale fermentations using biofilm operation are summarized here:
1. Biological waste water treatment in trickling filters (percolating filters or
fixed bed reactors) and in rotating disk fermenters. A system somewhat
related to trickling filters is the submerged filter, which, however, is con-
siderably limited in application as only two-phase systems can be handled
(LIS processing). Gaseous substances must be dissolved in the L phase
before entering the filter.
2. The FBBR applied in biological waste water treatment (e.g., Atkinson,
1980), enzyme technology (e.g., Coughlin et aI., 1975), and some fermenta-
tions (Baker et aI., 1980). Higher values of mass transfer are attained
together with increased interfacial areas (aLls ,..., 3 '10 3 m 2 /m 3 ) and biomass
concentration (x ,..., 40 kg m- 3 ), uniform distribution of solids, and so on
(Baker et aI., 1980).
3. The old "quick" vinegar process using packings of beechwood chips as
support material for cells to increase the area of contact with the feed liquor
trickling through.
4. Animal tissue culturing, in which cells are growing adhering to a surface in
the presence of un stirred layers of medium (trays).
6.8 Dialysis and Synchronous Culture Operation 371
5. Bacterial leaching of ores for the recovery of metals by natural adhesion on

solid materials.
6. In conventional STRs. Such STRs also imply biofilm formation (wall
growth), a fact that is often ignored.
7. In natural aquatic systems, water distribution systems, waste water pro-
cessing, heat exchangers, fuel consumption by ships, and even human
disease. It should be noted that biofilms are emerging as a critical factor in
these. The most common method of controlling biofilm accumulation in
practice is chlorination, a process that is limited, due to its toxicity. Thus,
research in biofilm processing has been stimulated (Charaklis, 1981). Often
the term "fouling" is used to refer to the undesirable formation of deposits
on surfaces that impede the flow of heat, increase the fluid frictional resis-
tance, increase the rate of corrosion, and result in energy losses.
6.8 Dialysis and Synchronous Culture Operation

Conventional continuous or batch culture methods suffer from some restric-
tions that are disadvantageous for technical or economic reasons (e.g., accu-
mulation of toxic and/or inhibition of metabolic end products, limited cell
densities, etc.), or are disadvantageous for scientific research purposes (e.g.,
cultivations are related to the statistical mean of the cell population and not
to the individual cell). Recent extensions of the continuous culture methods,
that is, dialysis and synchronous culture techniques, are able to fill this gap.
6.8.1 DIALYSIS (MEMBRANE) REACTOR OPERATION

6.8.1.1 Potentialities
There are many different kinds of membrane processes, but all have certain
features in common. In all of them, a fluid containing two or more components
is in contact with one side of a membrane that is more permeable to one
component (or a group oflike components) than to other components-it is
a selective membrane. The other side of the selective membrane is in contact
with a fluid that receives the components transferred through the membrane.
To cause the transfer of components, there must, of course, be a driving force
of some kind. Such a force may be a transmembrane difference in concentra-
tion, as in dialysis; electrical potential, as in electrodialysis; or hydrostatic
pressure, as in reverse osmosis, ultrafiltration, and microfiltration. The free-
dom to choose both the driving potential to be used and the diffusing species
(solute or solvent) has led to a multitude of nomenclatures for membrane
processing. Generally, processes involving the diffusion of solvent are termed
osmosis or ultrafiltration, and those for the solute are termed dialysis.
Dialysis is among the oldest of the membrane processes, but it has found
only a few industrial applications. However, greater industrial usage may
develop for the newer processes of Donnan dialysis and ion-exchange dialysis
(see Lacey, 1972).
In dialysis, that is, in the separation of solute molecules through their
unequal diffusion through a semipermeable membrane because of a concen-
tration gradient, the small-molecular-weight products are removed from the
immediate environment of the bacterial cell (and ultimately from the intra-
cellular enzyme site), which relieves feedback inhibition by a product that
normally regulates its own production. As more product is withdrawn by
dialysis, more substrate is consumed and more product is made: The fermen-
tation thus becomes more efficient. As the cell population attains high density,
the substrate is increasingly converted into product by maintenance rather
than by growth metabolism, also thus improving fermentation efficiency. On
the other hand, the application of dialysis increases costs, complicates opera-
tion, and eventually reduces efficiency because of membrane fouling. The
potential advantages must compensate for the disadvantages if a dialysis
process is to be useful.
Although membrane processes have been studied for more than a century,
they have only recently become of interest for industrial separations (e.g.,
Applegate, 1984). This interest stems primarily from a basic advantage of
membrane processes: They allow separation of dissolved materials from one
another or from a solvent, with no phase change. Since the energy cost
represents a sizeable portion of the total operating cost for most separations,
the possibility of effecting worthwhile economics in energy cost by using
membrane processes is attractive (e.g., Gerstenberg et aI., 1980; Pye and
Humphrey, 1979; Rautenbach and Albrecht, 1982). However, for total operat-
ing cost to be reduced, flux through the membranes has to be great enough
to permit use of reasonably small areas of membrane, resulting in low equip-
ment costs. Fermentation operation with dialysis culture systems was one of
the first practices of use of cell retention. Laboratory equipment for cultivation
of microorganisms with the removal of cell-free medium is a useful tool in the
study of continuous culture.
6.8.1.2 Principles and Performance Equations

Four basic modes are generally possible in dialysis culture operation (Fig.
6.47): (a) continuous reservoir and continuous fermenter, where Fros and
Fcerm > 0; (b) batch reservoir and batch fermenter, where Fres and Fcerm = 0;
(c) batch reservoir and continuous fermenter, where Fres = 0 and Fcerm > 0;
and (d) continuous reservoir and batch fermenter, where Fres > 0 and FCerm =
O. The quantification of membrane performance is achieved with the aid of a
formal approach based on Fick's law of diffusion using membrane permeabil-
ity coefficient Pmb' which includes several unknown factors that must be
measured experimentally:
(6.146a)
Fres 'term
5
x p
DIALYZER
p
FIGURE 6.47. Typical experimental setup of dialysis culture using a reservoir for
substrate feed S, a fermenter for bioconversion, and a chamber with a semipermeable
membrane M to retain biomass X.
where ns = rate of permeation [kg/h]

Amb = membrane area [m 2]
As = substrate concentration gradient across the membrane
Pmb [kg/kN h] can be estimated from a plot ofL flux (nS/Amb) versus pressure
on the feed side, according to
ns
- = J 1 = Pmb(Ap - An) (6. 146b)
Amb
Here An is the osmotic pressure offeed (permeate). The flux rate of the solute
(salt) obeys Fick's first law, that is,
(6. 146c)
which normally is heavily influenced by concentration polarization. Thus, the

solute flux is proportional to the solute difference inside the membrane (AC2r,
which in case of concentration polarization is substituted by cT = K 2 c2 ,
leading to an expression instead of Equ. 6.146c by using the bulk solute
concentration difference
(6}46d)
The performance equation of a membrane can be derived by introducing

intrinsic rejection (R j = 1 - c~/cT) and observed rejection (Robs = 1 - C~/C2)
with c~ being the bulk concentration at the permeate side:
In 1 - Robs = 25~ Re1/4. SC2/3 + In 1 - Rj (6. 146e)

Robs U Rj
where v is the permeate velocity and u is the feed velocity. A plot of (1 - Robs)!
Robs versus JdUO. 75 will give a straight line, with an intersect proportional to
Rj ( "'cn
The permeability coefficient reflects the overall resistance of the composite

barrier, and therefore includes external transport limitations (cf. Sect. 4.5.1),
which can be quantified using the two-film theory (cf. Equ. 3.30):
_1_=~+ I1 (6.147)
Pmb Pmb i D
where P~b = true permeability
D = diffusion coefficient
bi = thickness of transport-limiting liquid films.
Thus, the rate of dialysis is strongly dependent on the liquid velocity at the
membrane surface u. The thicknesses of the liquid films decrease as the bulk
velocity near the membrane increases (see Equs. 4.1a and 6.120).
Mathematical analysis always consists of the writing of suitable balance
equations by the insertion of kinetic terms. In the case of dialysis culture,
balances must be made either on the reservoir chamber or the fermenter
chamber. With use ofEqu. 6.146a for membrane permeation, the basic balance
equations for dialysis culture are as follows:
The balance in the reservoir in general form is
Accumulation = input - output + permeation
which gives for substrate and product (and not for cells)
(6.148)
and
(6.149)
A similar balance can be established in the fermenter for substrate, biomass,

and product in general form:
Accumulation = input - output + permeation reaction
and
(6.151)
and
These five mass balance equations plus model equations for kinetics (rx, rs ,
FIGURE 6.48. Calculated comparison of dialysis and x.S

nondialysis continuous culture with respect to concen- A
tration of biomass x and substrate s (a), productivity
D x (b), and process efficiency expressed as produc-
tivity ratio (c). Parameter values are: Vrerm = 11, I1mBx = J
I
1 h-l, Ks = 0.2 gl-l, YXIS = 0.5, s?erm = s?es = 10 gl-l, I
Fres = 0.51h-1, Pmb' Amb = 420 cm 3jh. (Adapted from 10 SNON DIALYSIS I
Schultz and Gerhardt, 1969.) I
I
I
I
I
5 I
I
0 a
0 0.5 1-0
O,XACTUAl
O.X
A
10 O,XIDEAl
%
100 --------.,
NON DIALYSIS
......"
A \
\
5 50
b
c
0
0 0.5 0 0.5 1 -0
and rp; see Chap. 5) are mathematically sufficient to determine dialysis culture
operation (Coulman et at, 1977; Schultz and Gerhardt, 1969).
Case (a) of completely continuous dialysis operation is the only one that is
essentially steady state in nature, so that Equs. 6.148 through 6.152 can be
simplified as the time derivates are all equal to zero. Solutions are fully
described in original papers, together with the behavior of x, s, and the critical
dilution rate Dc as a function of dialysis. With this solution, a comparison
between dialysis and nondialysis continuous culture can be realized, as shown
in Fig. 6.48. The comparison is carried out with respect to cell concentration
(a), productivity (b), and efficiency of cell production (c), which is defined as
the actual production rate divided by the production rate equivalent to
complete utilization of the substrate supplied.
The most striking and important difference between nondialysis and dialy-
sis continuous culture is the much higher cell concentration attainable in
dialysis culture, especially at low dilution rates. Further, it is seen that the
maximum production rate in continuous dialysis culture is achieved at a lower
dilution rate than in nondialysis continuous culture. This effect, however, is
reached at the expense of lower efficiencies in converting substrate to cells.
Case (b) of completely batch operation of dialysis is the operational mode
that is still predominantly used in process analysis. Mathematically, the values
376 6. Bioreactor Perfonnance: Process Design Methods
EXPONENTIAL LI NEAR GROWTH

GROWTH
I
I
I
I
I
I
I
......... I
-....
sferm i -----5 res,1
FIGURE 6.49. Expected changes in cell mass concentrations for dialysis culture systems
operated with batch fermenter and batch reservoir (X2 and Sres,2) or batch fennenter
and continuous reservoir (Xl and sres, d, The critical time when substrate diffusion
through the membrane becomes limiting is indicated as t erit Exponential, linear, and
asymtotic growth can be observed, (Adapted from Schultz and Gerhardt, 1969,)
of Fres and Fferm in Equs. 6,148 through 6.152 become zero, The expected
growth pattern for the fully batch dialysis culture is best evaluated by com-
puter and is shown in Fig. 6.49. The growth cycle exhibits two separate phases:
exponential growth until the region of S limitation, and linear growth by
S-diffusion rate limitation. With dialysis culture, the exponential growth phase
is extended because of the additional nutrient via the membrane.
The case (c) of batch fermenter/continuous reservoir dialysis culture can be
evaluated similarly. The behavior is also shown in Fig. 6.49, and the properties
are analogous to those in case (b). Linear growth can be maintained, because
Sres is constant due to membrane diffusion.
The limitations of dialysis fermentation are evident, since the inherently

slow process of diffusion of both nutrients and metabolic products through
the membrane is the rate-determining step. Inhibitory and/or toxic substances
can accumulate and limit maximum cell concentration.
In the case of product formation in dialysis culture, the only additional
information needed (see Equs. 6.148~6.152) is the permeability characteristics
of the product through the membrane material and the kinetic model for
product formation (see Sect. 5.4). The intricacies of fitting a mathematical
model to the kinetics of product formation were illustrated for steroid conver-
sion (Chen et aI., 1965), in which phenomena such as substrate solubility,
co-precipitation, feedback, and substrate inhibition have to be taken into
account.
Analytical solutions are possible for particular situations, for example, in
fully continuous operation. Here the final concentration of product that
cannot diffuse across the membrane will be P = P/Vcerm' However, if the

product is diffusible (Pmb,P > 0), then it will be distributed between the fer-
menter and the reservoir, P = P/(Vcerm + v..es). Thus
Pferm (6.153)
Pres = Fres/(Pmb.P' A mb ) + 1
To obtain a large fraction of product in the reservoir effiuent free from cells,
both the permeability Pmb,p' Amb and F res must be large in comparison with
Fferm . This type of operation results in a low concentration of product, which
creates difficulties in product recovery.
From the preceding analyses it is apparent that the behavior of reactor
systems is more closely linked to the operation of the fermenter chamber than
to operation of the reservoir. Therefore, the first step in designing a dialysis
culture is to choose between the security and flexibility of batch cultures and
the uniformity and economy of continuous techniques.
6.8.1.3 Membrane Reactors and Applications

Dialysis processes were originally developed as simple flasks with internal
filtration (Gerhardt and Gallup, 1963). Later, modifications (e.g., Dostalek and
Haggstrom, 1982) and modern designs (rotating microfilter, Sortland and
Wilke, 1969; the rotorfermenter, Margaritis and Wilke, 1978a,b; hollow fiber
reactors, e.g., Kan and Shuler, 1976, 1978) were developed. Some reviews
should be mentioned (e.g., Applegate, 1984; Meiorella et aI., 1981; A. Moser,
1985a) in which different reactors are compared and discussed in terms oftheir
fields of application (fermentation and enzyme technology).
In several microbial processes, economic efficiency can be drastically
increased when a technical solutions to the problems of discharging accumu-
lated inhibitory or toxic products or metabolites, or of using substrates
consisting of dispersed solids, can be found. Examples are the production of
ethanol from carbohydrates (Charley et aI., 1983; Kosaric et aI., 1980; Rogers
et aI., 1980), utilization of whole or deproteinized whey for the production of
food yeasts or nitrogenous feed supplement for ruminants (Gerhardt and
Gallup, 1963, Lane, 1977), production of salicylic acid from naphthalene
(Abbott and Gerhardt, 1970a), threonine biosynthesis (Abbott and Gerhardt,
1970b), and continuous aseptic production of phytoplankton (Marsot et aI.,
1981).
A number of alternative approaches are available as technical solutions for
the above-mentioned problems (e.g., Hamer, 1982; Lilly, 1982; Meiorella
et aI., 1981): vacuum membrane retractive and extractive fermentation, and
biphasic processing (Mattiasson, 1983; Reisinger et aI., 1987) and ion-
exchange resin culture techniques. Experimental tests of dialysis culture pro-
cesses are described in the literature for ammonium-lactate fermentation of
whey (Stieber et aI., 1977), and an improved mathematical model was later
developed (Stieber and Gerhardt, 1979) incorporating P inhibition. Recently,
the mathematical model was modified to incorporate a second feedstream of

cells, substrate, and product into the fermenter. The behavior of this dialysis
culture system with a cell-feed flow from a prefermenter or from cell recycling
was examined and the process was improved (Stieber and Gerhardt, 1981a,b).
Results under steady-state conditions showed that these dialysate-feed sys-
tems are a new and useful way to immobilize living cells to produce a
metabolite at a high rate for a prolonged time. The substrate consumed by
the cells is converted to product via maintenance metabolism only and is
sterilized by dialysis. Similar considerations in the case of the rotor fermenter
(RF) led to an expression comparing cell productivity in this membrane
reactor with that received in a CSTR. The cell productivity (cf. Equ. 6.9) for
the rotor fermenter and CSTR at the same level of s and assuming the same
Y value is expressed as a ratio
r F,
~=1+~ (6.154)
rX,CSTR FB
where FB is the volumetric cell bleed rate [ljhJ in the rotor fermenter and FF
is the filtrate flow rate [ljhJ.
As shown by Equ. 6.154, high values may be obtained in the RF by
employing a high ratio of feed to bleed rate. Thus, by suitable choice of flow
rates, the RF may fulfill the functions of both an ordinary fermenter and a
centrifuge cell separator (Margaritis and Wilke, 1978b). With similar argu-
ments, an expression can be derived giving the productivity ratio for ethanol
as product:
rp,RF _
----
1+-
FF.-PRF
- (6.155)
rp,CSTR FB PCSTR
As seen from Equ. 6.155, if PRF ~ PCSTR and FF ~ FB, then the productivity
ratio Rp > 1. Thus, by suitable choice of operating conditions for FF > FB so
that FF is maximized while PRF > PCSTR is maintained, about 10 times greater
ethanol productivity was experimentally found in the rotor fermenter from a
CSTR, thus illustrating the advantage of membrane reactor operation.
6.8.2 SYNCHRONOUS CULTURE OPERATION

Contemporary knowledge of cell metabolism is based on studies of asyn-
chronous populations that are often of poorly defined origin and from in vitro
rather than in vivo studies. As an alternative, it might be worthwhile to think
of growth and metabolism in terms of the cell cycle and post-cycle patterns.
It thus becomes possible to relate presently incompatible results from batch
and continuous cultures. For instance, if we consider the culture in terms of
the cell, the growth curve can be rearranged to bring out a possible connec-
tion between growth and secondary metabolism, which Bu'Lock (1965) and
Bu'Lock et al. (1965) described as the "tropho- and idiophase" (see Dawson,
1972a).
For many research purposes it is of great importance to have a synchronous

culture of microorganisms. Such situations appear with, for example, muta-
tion experiments, and are also of general interest in respect to some inade-
quacy of continuous culture theory. Results from a CSTR at steady state relate
to the cell in terms of dilution rate, which represents a reasonable, yet average,
value of the homogeneous proliferation that exists in the single-stage CSTR,
but which is not necessarily a true reflection of microbial behavior. In recent
years the advent of cell synchrony as a refinement in continuous techniques
has added new dimensions to the study of microbial growth (Cameron and
Padilla, 1966; Zeuthen, 1964). The potential advantage of synchronous culture
is directly related to the fact that in this technique the cells of the population
are all in the same stage of cell development (physiological state), doing the
same thing at the same time. The population serves as an amplification
of the cell and thus it becomes possible to study cell behavior by observing
the behavior of the population. Different methods are known, described as
synchronous/synchronized pulsed and phased culture techniques (Dawson,
1972a). Table 6.1 summarizes the characteristics of the various fermentation
techniques, which are illustrated in Fig. 6.50.
In "phased" culture, the cell population is synchronized, and growth rate,
nutrient supply per cell, popUlation numbers, and temporal activities become
experimental parameters routinely involved in the operation of the technique.
Phased cultures may be examined systematically like steady states in the
chemostat, but instead of the single-point (average) determinations of the
latter, patterns of cycle period ("cell cycle") activity that portray the replicative
performance of the cell during the doubling time period are obtained. Phased
culture permits us to rationalize, in a homogeneous population, the perfor-
mance of the cell under experimentally controlled conditions that can, if
necessary, be repeated indefinitely or varied at will (Dawson, 1980b). Dawson
(1972a,b, 1980a,b) has reviewed some of the significant results obtained with
phased cultures on the basis of cell cycle and post-cycle activities, an analysis
that the experimental dimensions of the chemostat cannot approach.
TABLE 6.1. Main characteristics of techniques available for cultivating

microbes.
Type Method System Growth rate Technique
Asynchronous Batch Closed Changing Traditional
"Bactogen"
Continuous Open Constant Chemostat
Turbidostat
Synchrony Batch Closed Changing Synchronous
Synchronous
Continuous Open Constant Pulsed
Pulsed
From Dawson, 1972a.

t& ~
~ DCSTR
w
TRANS I tNT GROliTrl
I~I
CSTRc X
STEADY STATE
1
t
1 ~t VVVlI
SYNCHRONY c
(I--
"PHASED
CULTURE"
ICb Q
lcb STEADY STATE CELL CYCLE
'lC:l
REPEATED
CONDITION
b
a c
FIGURE 6.50. Experimental setup for obtaining asynchronous cultures (DCSTR and
CSTR) and synchronous (phased) cultures (a), characterization of overall growth
conditions (b), and illustration of growth of individual cells (c). A comparison of the
different approaches shows clearly that in the DCSTR and CSTR, cells in the popula-
tions are randomized at all stages, while cells adjust their growth rate to the dosing
interval and become synchronized to it in the "phased culture technique" (Dawson
1980a).
Two general classes of methods exist for attaining synchrony in batch

and/or continuous culture:
1. Methods whereby cells at the same stage of division are selected or sepa-
rated from a randomly dividing population (usually by a mechanical or
physical procedure)
2. Methods in which the entire population is manipulated by applied con-
straints of the environment (which may be nutritional, physical, physiologi-
cal, or growth inhibitory) to align the cells and produce the inoculum for a
synchronized culture
A modified method was proposed by Kjaergaard and J0rgensen (1979) and
is probably better than others. The principle of the apparatus is that of a
normal chemostat, with the exception that the substrate fed with the dilution
rate contains no carbon source and that the carbon source is fed automatically
to the vessel at certain intervals in volumes that are negligible compared with
the reactor volume. The background to this procedure is that cells grow only
6.9 Integrating Strategy as General Scale-Up Concept in Bioprocessing 381
when a carbon source is present. As soon as the carbon source is added, the
cells start to assimilate and cell concentration increases according to
x = xo exp[(J.t - D)t] (6.156)
The amount of carbon source is adjusted in such a manner that only one
doubling in biomass can take place, which means that the sugar must be
consumed after the time In 21J.t.
The time period with no growth must be sufficient that x again is brought
to x o, which means that
x = Xo exp [ (1 - ~ }n 2] exp( - D . t) (6.157)
which gives
t= (~-;}n2 (6.158)
The total time between two additions of carbon source is therefore
ln2 + (~_ ~)ln2 = ln2 (6.159)

J.t D J.t D
This system will give a continuous synchronous culture in which the cell
growth rate is independent of dilution rate (In 21D) as long as D < J.t. A new
generation of cells is produced every In 21D.
For microbial biomass production phased cultures would appear to have
a 50% production advantage by permitting the culture volume to be harvested
at doubling time intervals, rather than at replacement or retention time
intervals (which correspond to approximately 1.5 times the doubling time).
6.9 Integrating Strategy as General Scale-Up Concept

in Bioprocessing
The integrating strategy has already been presented in Chap. 2 as a funda-
mental approach for simple but adequate bioprocess modeling (cf. Fig. 1.4).
Especially in Fig. 2.13, the integrating point of view was represented by a block
diagram of a bioreactor, indicating the problems involved. Agitation (micro-
mixing), macromixing, and aeration will influence the internal and external
transport phenomena that ultimately determine the extracellular environ-
ment. Simultaneously, this environment is modified by the feeding of sub-
strate, which is used for production of biomass and products. An important
aspect in this scheme is the feedback from biomass and product to the
environment. It is well known that most secondary metabolites are produced
with filamentous microorganisms, and the properties of the broth can drasti-
cally change. All of the transport phenomena of the bioreactor will depend
ultimately on physical properties of the fermentation fluid, including viscosity,
shear gradients, density, gas holdup, and so on. Thus, the integrating strategy
deals with the joint aspects of balancing, stoichiometry, thermodynamics, and
kinetics of reactions and transports.
6.9.1 STOICHIOMETRY (BALANCING METHODS) ApPLIED IN

BIOPROCESS DESIGN
As a consequence of the increased worldwide competition in bioprocessing
compared with chemical processing, considerable attention has been directed
toward improvement in the efficiency of technical bioprocesses. In this situa-
tion, formulating balanced stoichiometric equations is a powerful technique.
Taking a general case of (cf. Equ. 2.8):
VS(CaHbOJ + vNNH3 + Vo' O 2 -+ vx(CaHpOyN{,) + vp(Ca,HfJ,Oy,N~,)
+ vW'HzO + Vc C0 2
the elemental balances are (cf. App. I):
C: a . Vs = rx' Vx -+ rx' . Vp + Vc
H: + 3 . vN = [3. Vx + [3" vp + 2 . Vw
b . Vs
0: c'vs + 2'vo = Y'v x + y"vp + Vw + 2've
N: vN = (j . Vx + (j' . Vp
Generally, there are two many unknowns (e.g., 15). In case the composition
of S, X, and P are known (a,b,c; rx,[3,y,(j; rx', [3',y', (j') there remain only six
unknowns in four equations, which can then be solved using supplementary
information, for example, on the basis of gas analysis (Oz, COz)'
It was mentioned earlier that if the elementary analysis of the cells is known
(i.e., the stoichiometric coefficients in Equs. 2.1 and 2.8), then the mass balance
equation may be calculated without actually measuring the oxygen used or
COz formed. In this way, oxygen uptake and growth yield can be theoretically
predicted (Roels, 1980a). A macroscopic analysis based on elemental and
energy balances successfully describes relationships important for bioprocess
design. The existence of limits to the oxygen and substrate yield factors is
shown by Heijnen and Roels (1981) on this basis. For substrates of low degree
of reduction, the energy content of the substrate poses a limit to Yx1s . For
substrates of a high degree of reduction, a limit is posed by the carbon
available in the substrate. Simple models for the energetics of growth on
substrates with differing degrees of reduction are derived (Roels, 1980b), and
a quantitative description of growth on mixed substrates has been given
(Geurts et aI., 1980) showing a relationship between YATP and the PIO ratio.
Generally there are two points of interest. The first is, what effect does the
conversion yield have on the cost of production? The second is, how much
room is there for improvement; how do actual yields compare with theoretical
values?
CO2 Hp
FIGURE 6.51. Diagrammatic representation of a steady-state bioprocess in balance area

(reactor) following the macroscopic principle by analyzing elemental composition of
significant process variables (substrate, nitrogen source, biomass, product, 2 , CO 2 ,
H 2 0). (Adapted from Roels, 1980a.)
On the basis of the preceding analysis of writing mass and energy balances
(cf. Sects. 2.2.3 and 2.4.2, as outlined in Fig. 6.51 together with Appendix I),
there exist a number of interesting papers in the literature. In the absence of
information on the biosynthetic pathway, an alternative approach based on
reaction stoichiometry was applied to penicillin synthesis (Cooney, 1979;
Cooney and Acevedo, 1977). The theoretical maximum yield of penicillin
from glucose was calculated; the actual conversion yield could be improved
substantially by increasing and sustaining the specific rate of penicillin pro-
duction and by minimizing maintenance metabolism. The actual yield of
penicillin from glucose was shown to be an order of magnitude lower than
the theoretical value. Increasing efficiency of glucose utilization for penicillin
will markedly decrease the oxygen demand. This is important not only to
reduce production costs, but also to increase the capacity of existing equip-
ment, since it is often the OTR ability that is rate limiting in production.
Applying the strategy in formulating a simple unstructured model accord-
ing to Fig. 2.19, Heijnen et al. (1979) developed elementary and enthalpy
balances for the penicillin production process. This macroscopic analysis
used carbon, hydrogen, nitrogen, oxygen, sulfur, phosphorus, and enthalpy
balances for 11 relevant process variables (compounds), so that according to
the element-species matrix (cf. Appendix I), at least five kinetic equations are
needed to complete the model. On the basis of this fermentation model, the
authors presented a series of simulation studies and showed some interesting
results. They emphasized that a growth-coupled penicillin production pro-
vides an adequate description of most of the observed phenomena, which has
often led to the assumption of non-growth-associated or age-dependent peni-
cillin productivity. There appears to be no experimental evidence necessitating
introduction of an age dependence or a time delay in modeling the kinetics of
penicillin production.
Other important results concern the glucose balance and the effect of
glucose feed rate schemes in fed-batch cultures. It was shown by this analysis
that 20% of the glucose is used in the production of mycelial dry matter, 10%
is used for penicillin synthesis, and as much as 70% is used in maintenance
processes. In obtaining high penicillin yields, the glucose feed scheme is shown
to be of crucial importance. A scheme using an increasing feed rate of glucose
as a function of time is superior to use of a constant or a decreasing feed rate.
However, in the investigations of Heijnen et aI., only one point in each scheme
of feed rate was simulated, so a complete picture is not given.
Recently the impact of sugar-feeding strategies was investigated, and results
showed that maximum productivities are more or less independent of the
feeding scheme (Bajpaj and Reuss, 1981). A strategy proposed by Mou (1979)
was employed in which the biomass growth rate is controlled at one preset
value during growth phase flGr and at another preset value during production
phase flPr by supplying a rapidly metabolizing sugar. Thus, in principle, there
are three external parameters that must be set to define an experiment: flGn
flPn and Xtp which is the transition concentration of biomass and which serves
as an indicator for the onset of the production phase. During the fast growth
period, empirical correlation using CO 2 production data accurately reflects
the cell concentration to 1 gil and the instantaneous value of fl to 0.01
h- 1 :
Xt = Yx1c ' J: rc' dt [g cells] (6.160)
Since the cells are growing at a low growth rate during the transition and
production phases, maintenance activity and endogeneous metabolism are
significant, and it is not possible to use Equ. 6.160. Instead, overall and
instantaneous carbon-balancing equations allow successful on-line calcu-
lation of the cell concentration and instantaneous growth rate (Mou and
Cooney, 1982a,b):
(6.161)
where Cs, CPAA ' Cco 2 , Cp , and Cx are carbon content in substrate consumed,
phenylacetic acid as penicillin precursor fed, CO 2 , penicillin, and cells. These
equations, together with a feedback control method, enable the computer to
control the production phase growth rate with minimum error (0.002 h -1).
The examples presented have shown that the construction of simple
unstructured models based on some notion of relevant kinetic equations in
combination with the very important concept of elemental composition and
enthalpy balance is of great help in understanding factors relevant in the
optimization and control of even such complex fermentation processes as the
production of antibiotics.
Another important field of successful application of the balancing method
is biological waste water treatment. A unified approach toward mathemati-
cally describing activated sludge waste water treatment and to describing the
biokinetic and stoichiometric relationships involved, has led to many signifi-
cant insights into the principles of treatment process designs and operation
(Sherrard, 1977). Determination of the quantities of nutrients that must be

added to nutrient-deficient waste water is important in the successful design
and operation of treatment facilities. Nutrients must be added to provide
balanced growth media and to avoid algal growth (in the case of excess
nutrients) or incomplete conversion or takeover of filamentous bacteria (in
the case of insufficient nutrients). Stoichiometric equations may be obtained
at any given value of mean residence time so that the stoichiometric coeffi-
cients of all components are found and the requirements of oxygen, nitrogen,
and phosphorus can be determined (Sherrard and Schroeder, 1976). Beyond
these predictions of requirements for inorganic nutrients and 02' the sludge
production and effluent quality on further treatment processes can be rigor-
ously analyzed as a function of the mode of process operation (Sherrard, 1980).
A similar fundamental procedure of developing a reaction scheme, handling
kinetic expressions, and reducing the system to important reactions and
components was applied to the case of biological nitrogen removal (Irvine et
aI., 1980).
6.9.2 INTERACTIONS BETWEEN BIOLOGY AND PHYSICS VIA

VISCOSITY OF FERMENTATION MEDIA
The rheological classification of liquids is normally given by a general expres-
sion, where the shear stress applied, [force . length -2] and the resulting shear
rate y [time- 1 ] are correlated as follows (e.g., Metz et aI., 1979):
1: = 1:0 + K(yr (6.162)
where K = power law constant or consistency index
'0 = yield stress
m = power law index or flow behavior index
This is known as the power law.
The following rheological classification of fluids is possible (Fig. 6.52):
FIGURE 6.52. Rheogram of fermentation fluids in

the form of shear stress ! versus shear rate y in
case of Newtonian and non-Newtonian fluids (cf.
Equ.6.162).
1. Newtonian fluids, with m = 1 and To = O. Thus, from Equ. 6.162

T=I]Y (6.163)
where 1] is the viscosity (dynamic) [Ns m- 2 ].
2. Non-Newtonian fluids, of which there are several cases:
a. Power law fluids, where To = 0
T = K(yr (6.164)
with m < 1 for pseudoplastic fluids and dilatant liquids with m > 1.
b. Bingham plastic fluids, where
T = To + 1]. Y (6.165)
The Casson equation is sometimes a very successful alternative approach:
(6.166)
where Kc is Casson viscosity [(N sm- 2)1/2]. This equation is often preferred
to the power law equation in the case of mycelial suspensions (see Roels et aI.,
1974). Here, apparent viscosity is a function of shear rate:
(6.167)
It can be shown that many fermentation media behave like non-Newtonian

fluids. Thereby (apparent) viscosity can be a function of either Sand/or P
and/or X concentration, exhibiting Newtonian and non-Newtonian behavior.
The estimation of the rheological properties of biological fluids is summarized
in the literature (e.g., Charles, 1978; Metz et aI., 1979). Reuss et al. (1980)
incorporated rheological considerations into process modeling. When visco-
sity is taken into account, clearly any correlation that does not contain the
variation of viscosity during the fermentation must fail. An approach recently
developed by Zlokarnik (1978), originally proposed for water and water-salt
solutions, can be successfully used for this purpose (cf. Equ. 3.81):
VL PG
kLl a- =f 2/3 (6.168a)
FG FG pdv g)
The first group (NoTR ) combines kLl . a with liquid volume and gas flow rate
FG; the second group includes power of the impeller PG, liquid density Pu FG,
kinematic viscosity (v = I]/g), and acceleration due to gravity g. As can be seen
from Fig. 6.53, this correlation is adequate for experimental observations of
kLl . a during fermentations of Penicillium chrysogenum and Aspergillus niger.
The viscosity of fermentation broth in this case was shown to be primarily a
function of biomass concentration. Recently, an alternative equation for this
purpose (Stiebitz and Wolf, 1987) was derived, which is superior to Equ. 6.168a
and which is valid in the range 2 < I] < 16 m Pa s (up to 50% glucose
concentration in H 2 0):
--
-- --
__ -- PENICILLIUM CHRYSOGENUM
-- ASPERGILLUS NIGER
FIGURE 6.53. Application of chemical engineering correlation for oxygen mass transfer
according to Equ. 6.168a (cf. Equ. 3.81) to technical bioprocessing in case offermenta-
tions with Penicillium chrysogenum and Aspergillus niger under various conditions. The
dotted lines indicate observed deviations in the bioprocess. (Adapted from Reuss et
ai.,1980)
kLl a VL (V2)1/3 = c1[ PG ]C2 (6.168b)

FGHo 9 FGPLHog
with C 1 = 5.9 10- 5 and C 2 = 0.52.
Equation 6.164 was used where the shear rate y in the bioreactor was
estimated from a known correlation with agitation speed n in a STR (Calder-
bank and Moo-Young, 1959; Metzner, 1957):
Yave = kn n (6.169)
with kn a constant (being about 10).
Reuss et al. (1980) showed then that the consistency index K, or 1'fapp' as
well as the flow index m in Equ. 6.164 was correlated with the biomass
concentration x following the empirical expressions
(6.170)
and
(6.171)
K
RESP.
llapp
PEN I C I LLI UM CHRYSOGENU~1
A
10.8
FIGURE 6.54. The correlation acc. to Fig. 6.53 can be verified by using a concept of the
dependence of apparent viscosity 1]app on shear rate y as a function of biomass concen-
tration according to Equ. 6.170. (Reuss et al. 1980).
Results of comparison of different measurement systems (turbine impeller and

helical ribbon impeller) are shown in Fig. 6.54 at varying concentrations of
P. chrysogenum biomass.
On the basis of this approach, the complete process model can be written
in the following equations for a fed-batch culture (cf. Sect. 6.2 and Bajpaj and
Reuss, 1980):
x dV
rx = f.1(x,s,o) - V'Tt (6.172)
l i s dV
rs = --f.1(x,s,o) - -qp(x,s, 0) - msx + Fs - - ' - (6.173)
YxlS YPls V dt
p dV
rp = qp(x,s, o) - k p , d' P - -'-d
V t (6.174)
and
1 1
--'"(xso)--'q(xso)-m x--
dV
y, f"'" Y. p" 0 V ' -dt+ k L 1 'a(o*-o)
XIO PIO
(6.175)
together with Equs. 6.168,6.170, and 6.171 for the interaction between visco-
sity and OTR and with kinetics according to
s 0
,u(x, s, 0) = ,umax K
s,x.x
+s K
o,x'x
+0 (6.176)
and
s 0'
q (X S o)-q .---- (6.177)
p " - P,max Kp + s(1 + s/K R ) Ko,px + 0'
where KR is the formal inhibition constant quantifying metabolite repression

for penicillin production (cf. Equ. 5.106).
This process model can be used to predict optimum productivity for peni-
cillin production, as depicted in Fig. 6.18. Productivity increases at low feeding
rates of sugar (S limitation), but decreases due to catabolite regulation (KR!)
at higher feeding rates Fs. The influence of kLl . a is demonstrated by the dotted
lines in the figure. If kLl . a is diminished as a consequence of increasing v due
to increased X, productivity goes down. This model was later used to search
for an optimum feeding strategy (cf. Sect. 6.2), and must be regarded as a
bioprocess model based on the interactions of transport and kinetics, where
the balance equations represent a general concept of scale-up instead of
conventional concepts based on power input, mixing times, Re numbers,
impeller tip velocity, or other factors of pure physical meaning. Recently, in
situ viscosimeters were developed, which fact will enhance the application of
the modeling approach described (Bjorkman, 1987).
6.9.3 INFLUENCE OF MYCELIUM-THE MORPHOLOGY FACTORS

("ApPARENT MORPHOLOGY")
Roels et al. (1974) have developed a model for the rheological behavior of
filamentous suspensions using the Casson equation by applying a measuring
technique consisting of observations of the torque exerted on a rotating
turbine impeller. Some considerations analogous to the rheological descrip-
tion of polymer solutions-the so-called "excluded-volume-concept"-lead
to the introduction of a morphology factor 0*, which is a function of yield
stress 'to and mycelial concentration X, defined as modified factor [Nm]:
(6.178)
The results, using data on P. chrysogenum, are shown in Fig. 6.55. As can be
seen, the power of X seems to be somewhat higher than 2, according to Equ.
6.178. Roels et al. (1974) found that the morphology index decreased during
batch fermentation. As a consequence of this approach, the parameters Kc
and 'to in the Casson equation can be expressed as dependent on 0*:
(6.179)
and
FIGURE 6.55. Relationship between yield stress '0 (cf. Equ.

1'0 6.162) and mycelial concentration x from different experi-
! mental measurements (., II) compared with Equ. 6.178).
(Adapted from Roels et aI., 1974.)
~_ 1/2( ~+Cl
C - 1Jo
1 ) (6.180)
v'o V (j*. x
From Equ. 6.166, thus
~ = ~'X(1 + Kc vr:f) (6.181)

Fa
where KclFa is to be substituted from Equ. 6.180.
Even though there are objections to the theoretical model (hyphae treated
as flexible chains forming spherical coils; network interaction between
branched hyphae; randomizing effect of Brownian motion essential for valid-
ity of polymer rheology theory), this formal approach is quite useful (Metz et
aI., 1979).
A semiautomatic method for the quantitative representation of mold mor-
phology was described (Metz, 1981) in which a variety of morphology indexes
were shown to be useful, for example, the effective hyphallength, Le; the total
hyphallength, L t ; and the hyphal growth unit, Lu (cf. Equ. 5.109).
The influence of engineering variables such as agitation and mean energy
dissipation 6 on the morphology of filamentous molds was also elaborated on
by the same group (van Suijdam and Metz, 1981), who concluded that an
increase in 8 gives a decreases in Le, L t , and Lu:
Le = C' 6- 0 . 25 (6.182)
Very recently, the concept of "apparent" or "engineering" morphology has
been presented (van Suijdam, 1986, van Suijdam and Dusseljee 1987), with
account taken ofthe macroscopic nature of microbial suspensions. The aim is
to have a parameter that
TABLE 6.2. Comparison of different "morphology factors" and approaches to bio-rheology.
Author Equation Equ.l1o Comments 0-,
\.0
van Suijdam and ry/~o = 1 + A1 F 'x 6.183a A1F = global morphology factor .....
::I
Dusseljee (1987) "apparent morphology"
ry/ryo = 1 + (A1l" x/I - s' x) 6.183b s = crowding factor
Einstein (1906; 1911) ry/ryo = 1 + 2.5 'l/Js 6.184 valid if l/Js < 0.05
~S
(JQ
Mooney (1951) In ~/1/o = 2.5l/Js (l - s'l/Js) 6.183d l/Js = volume fraction of solids IZl
rl
Vand (1948) ry/ryo = 1 + 2.5l/Js + 7.25# 6.185 valid if (Ps < 0.15 ...
2.5' , ~
(';
Eilers (1941) 6.186
t//ryo = 1 + 2(1 -l/Jslmox) ~
I'>
2 en
Shimmons et al. (1976) ~ = ryo'-- + 1.36l/Js 6.187
1 -l/Js ~
::I
1 (';
Reuss et aL (1979, 1982) 6.188 l/hs - mnx
.,
~/'Io = 1 - (h,' ,)1/2 e:..
IZl
Roels et al. (1974) c = p x 6.178 c = excluded volume
Metz et al. (1979) LI! 5.109 LII hyphae length ~
(';
van Suijdam and Metz (1981) Le 6.182 C

'0
Nestaas and Wang (1981) filtration characteristics: (")
1 o
PH = hyphae density == II ::I
g
'0
V specific volume of mycelia rl
dll = hyphae diameter

S
1:1:1
Vc = cake volume o
tfill .( KKC'1/filt ) PI!' VC 'g
(6.189) KKC = Kozeny-Carmen o
Vru,Vc = 32'A 2 'dP (I-Pu' Vd 3 ' dJ g
equation, factor en
en
tfill =
filtration time S
(JQ
Vc = specific cake volume
'IF = viscosity filtrate
dp = filtration pressure w
I,CJ
- is directly measurable
- has a meaningful relation to process variables
- is based on theoretical principles
- is independent in its dimensions from broth properties, especially from
biomass
In analogy to the Einstein equation, putting biomass x as a substitute for
volume fraction of solids (cf. Table 6.2 with Equs. 6.183-6.189), a more general
and flexible relation is achieved. The factor M F , thus, is the "apparent"
morphology, which is identical to the "intrinsic" viscosity or limiting viscosity
number
(6.183c)
being the substitute for the factor 2.5 in Equ. 6.184, which was a shape factor.
This approach, esp. Equ. 6.183b, is analogue to an equation (cf. Equ. 6.183d)
derived by Mooney (1951). M F , thus, can be estimated from the slope of the
plot of 11/110 versus x, respectively, 11 versus x. Deviations from the expected
straight line appear due to high volume fraction of biomass.
These morphology factors incorporated in the interaction model presented
(Equs. 6.172-6.177) and combined with kinetics and balancing methods will
form an integrating strategy for successful bioprocess analysis and design, as
shown in Fig. 2.15.
6.9.4 STRUCTURED MODELING OF BIOREACTORS (OTR)

It has been mentioned previously (cf. Sect. 3.3.2.3) that micromixing in high-
volume reactors can be handled adequately only by structured mixing models.
These concepts are observed in case of measuring mixing time in technical
scale units (cf. Equ. 3.19 in Table 3.1), but at the same time they exhibit
influences on other engineering parameters.
Many correlations are known for mixing time (e.g., Brown, 1981; Hugh-
mark, 1980; Joshi et aI., 1982; Reuss et aI., 1980), but these are valid, however,
only in bench-scale units.
An exception are recent data reported concerning gas holdup eG (Stenberg,
1984):
e ~ (~
P. )0.21.2
v
s (6.190)
G VL 1 + 41 Vs
power consumption
(6.191)
and O 2 mass transfer (with Equ. 6.190 for eG)
(6.192)
Thus, as a general conclusion it becomes clear that more sophisticated

mixing models are needed for reactor quantification, for example, OTR, on a
technical scale (cf. Fig. 3.10, 3.11, and especially Fig. 4.6 and 4.7).
Basically, two assumptions often used in structured (mixing) modeling are
of general importance:
- The reactor vessel can be divided into a number of well-stirred or plug flow
or even dead or scarcely agitated zones, called compartments, following the
"unit cell" approach.
- A certain flow rate of liquid FL or gas FG is recirculated between various
modules by the action of the impeller (pumping capacity of the stirrer Fp).
The general setup of structured models was graphically illustrated in Fig.
4.6 for a simple two-compartment model. According to the model presented,
mass balances over the reactor must be formulated fer significant process
variables (e.g., concentrations s, x, 0, p, hv, and c).
The unsteady-state balance of the nth module is given by the following
equation, written as a general expression:
Rate of change = rate of input - rate of output by flow
rate of consumption or formation rj (6.193a)
which takes the following form in case of mixed modules:
VN d;
dC' N
= F';n . Cj,N-l - Fox' Ci,N r j ' VN (6.193b)
with VN the volume of the nth compartment.

To determine the concentrations Cj leaving a set of perfectly mixed volume
compartments in series including loops it is necessary to apply Equ. 6.193 to
each compartment in sequence. This results in a set of coupled first-order
differential equations that are to be solved including kinetics and stoichio-
metric coefficients, and assuming initial conditions resembling the experi-
mental procedure in which an ionic tracer pulse is injected above the stirrer
at t = 0 (thus C = Co in the compartment just above the stirrer and c = 0
(elsewhere). At t = 00, the concentration is uniform (c = c*).
Depending on the number ofm6dules, the number of parameters increases.
A typical set of equations in the case of a three-compartment model, in analogy
to Fig. 4.6, will be of the following form including OTR:
Fp(c 2 - + kLa' V(cT -
cd cd - roo V = 0 (6.194a)
Fp(c 3 + C 1 - 2c 2 ) + kLa' V(c! - c2 ) - roo V =0 (6.194b)
Fp(C2 - c 3 ) + kLa' V(cr - C3) - roo V = 0 (6. 194c)
Dividing each term by Fp we obtain a set of equations with a dimensionless

parameter 1:, which is the ratio of two characteristic times (-r = tP/tOTR)' Here
tp is the mean internal residence time (V/Fp ), representing the pumping capabil-
ity of the stirrer, and tOTR = 1jk L a, representing the mass transfer capacity of
the stirrer. Analogous concepts of characteristics times and characteristic rate
constants are generally used in regime analysis (cf. Table 4.2, Sect. 4.2).
Similar structured approaches to mixing have been frequently used recently
in modeling OTR in production-scale bioreactors. Clearly, the dissolved
oxygen distribution in a reactor will depend on the interactions between
mixing and mass transfer. Figure 4.7 represented a simple unstructured model
for OTR in a stirred-tank reactor, taking into account the hydrodynamic
phenomena associated with the stirrer (Warmoeskerken and Smith, 1982).
Although this is a more mechanistic approach, the scale dependence is still a
problem. The influence of geometry is not clear. Therefore, structured model-
ing is needed for scale-up purposes, including a model for gas holdup eo, liquid
L-BALANCE1 ~----I \-------'1-\ L-BALANCE 2
FIGURE 6.56. Block diagram of the two-compartment model for oxygen transfer in a
production-scale bioreactor (ef. Fig. 4.6), including parallel work for mixed zone (1)
and bubble zone (2). For symbols see nomenclature. (Adapted from Oosterhuis, 1984.)
6.10 Final Note 395
circulation (to' CTD), the relation between gas bubble diameter dB and ku or
the coalescence and redispersion of bubbles (Oosterhuis, 1984).
Figure 6.56 gives an impression of the work related to this type of structured
modeling, showing the balances and auxiliary correlations for kLa, eG , P/Y,
Fp , and qQ, which are used to calculate the total oxygen transfer capacity of
the reactor (OTR) as combined from mixed zone and bubble zone (cf. Fig. 4.6).
Finally, the effect of structured models due to imperfect mixing in bio-
reactors upon product formation, yield, and conversion will be mentioned
here. It was shown by Oosterhuis (1984) that rp can be considered to consist
of two parts [ti = residence time in aerated compartment and tc = (Vi + V2 )/
Fp)]:
(6.195)
In the case of a relatively low exchange flow Fp between compartments 1 and

2, the ratio tdte can be used as a scale-up or scale-down criterion.
6.10 Final Note

In place of general conclusions, Fig. 6.57 summarizes the overall strategy of
bioprocess technology by graphical means. After the qualitative character-
istics of the biocatalyst/substrate system are worked out in a microbiological
laboratory, the first phase consists of quantitative measurements and mathe-
matical model building of biokinetics in a "perfect bioreactor" (Chaps. 4 and
5). Quantitative measurements and model building for various types of reac-
tors form the second phase (Chap. 3). Process development (Chap. 6) is
concluded only with the formulation of a process model made up of transport,
kinetics (including interactions), and scale-up elements based on additional
pilot plant measurements.
All of the methods presented in this chapter agree with the strategy given
in Figs. 1.4, 2.14, and 2.20. The necessary kinetic data come primarily from
measurements made on discontinuous processes as a first working hypothesis.
However, these data are not readily transferable to other modes of operation:
The data reflect a strong coupling between the reactor operation and bio-
logical behavior. Maintaining a systematic approach, as suggested by the
procedures described in this book, should lead step-by-step to a more reliable
method for planning bioprocess operations.
Acknowledgment. Most of the computer simulations in Chaps. 5 and 6 were

performed by the group of Prof. Dr. W. Knorre of ZIMET (Central Institute
for Microbiology and Experimental Therapy, Biotechnology Division) at the
Academy of Sciences of the German Democratic Republic in Jena (Prof. F.
BIOCATALYST SUBSTRATE
I "- /
PERFECT BIOREACTOR 1--- -- ---
.hW tJJ lJJ lbJ.... ~ f-
~\/~MAT'
---I ~10DEL :
r- . BIOREACTOR MODE'
. - L-..._ - - - - - -
!~ REACTOR
~ nj
--IPILOT PLANT REACTOR I-- ------- ----- ---

, ~
L __ -I PRODUCTION REACTOR 1
+
PRODUCT
FIGURE 6.57. Summary diagram of work flow in the systematic development of a

bioprocess of the presented integrating strategy. The diagram is based on the inter-
action between kinetics (Chap. 5) and transport (Chap. 3) processes, which are clarified
during a kinetic analysis (Chap. 4). As a special situation, the design and utilization of
new types of reactors are shown (discontinuous stirred vessel, DCSTR; bubble column,
BC; semicontinuous stirred vessel, SCSTR; recycle reactor, RR; continuous stirred
vessel, CSTR; continuous cascade, NCSTR; tower reactor, TR; continuous plug flow
reactor, CPFR; fixed and fluidized bed reactor, FBR).
Bergter), based on a scientific collaboration with the author's group at the

Technical University in Graz, Austria.
BIBLIOGRAPHY
Abbot, B.J., and Gerhardt, P. (1970a). Biotechnol. Bioeng., 12, 577.

Abbot, B.J., and Gerhardt, P. (1970b). Biotechnol. Bioeng., 12,603.
Adler, I., and Fiechter, A. (1983). Chem. Ing. Techn., 55, 322.
Agrawal, P., and Lim, H.C. (1984). Adv. Biochem. Eng., 30, 61.
Aiba, S., Humphrey, A.E., and Millis, N.F. (1973). Biochemical Engineering. New York
and London: Academic Press.
Aiba, S., et al. (1976). Biotechnol. Bioeng., 18, 1001.
Alper, E., et al. (1980a). Chem. Eng. Sci., 35, 217.
Alper, E., et al. (1980b). Chem. Eng. Sci., 35, 1264.
Andrews, J.F. (1971). Biotechnol. Bioeng. Symp., 2, 5.
Andrews, J.F. (1982). Biotechnol. Bioeng., 24, 2013.
6.10 Final Note 397
Andrews, J.F., and Tien, C. (1982). Amer. Inst. Chern. Eng. Journal AIChEJ, 28,182.
Applegate, L.E. (1984). Chern. Eng., June, 64.
Arcuri, E.J., and Donaldson, T.L. (1981). Biotechnol. Bioeng., 23, 2149.
Aris, R., and Humphrey, A.E. (1977). Biotechnol. Bioeng., 19, 1375.
Atkinson, B. (1973). Pure Appl. Chern., 36, 279.
Atkinson, B. (1974). Biochemical Reactors. London: Poin.
Atkinson, B. (ed.) (1980). Symposium on Biological Fluidized Bed Treatment of Water
and Wastewater, Water Research Centre, Stevenage Lab., Elder Way, Stevenagej
Hertfordshire, Great Britain, April 1977.
Atkinson, B., and Fowler, H.W. (1974). Adv. Biochem. Eng., 3, 221.
Atkinson, B., and Knights, A.J. (1975). Biotechnol. Bioeng., 17, 1245.
Atkinson, B., and Kossen, N.W.F. (1978). Proceedings of the 1st European Congress
on Biotechnology, Interlaken, Switzerland, Dechema Monograph, 82, 37
Atkinson, B., et al. (1980). Proc. Biochem., May, 24.
Atkinson, B., and Mavituna, F. (1983). Biochemical Engineering and Biotechnology
Handbook. Nature Press, Macmillan. Byfleet, Surrey England.
Atkinson, B., and Davis, I.J. (1972). Trans. Inst. Chern. Engrs. 50, 208.
Bahl, H., et al. (1982). Eur. J. Appl. Microbiol. Biotechnol., 15,201.
Bailey, J.E., and Ollis, D.F. (1977). Biochemical Engineering Fundamentals. New York:
McGraw-Hill, p. 642.
Bajpaj, R.K., and Reuss, M. (1982). Canad. J. Chern. Eng., 60, 384.
Bajpaj, R.K., and Reuss, M. (1981). Biotechnol. Bioeng., 23, 717.
Bajpaj, R.K., and Reuss, M. (1980). J. Chern. Techn. Biotechnol., 30, 322.
Baker, C.G.J., et al. (1980). In Moo-Young M., et al. (eds.). Proceedings of the 6th
International Fermentation Symposium, Vol. 1. Oxford: Pergamon Press, 635.
Barford, J.P., et al. (1982). In Bazin, M.J. (ed.). Microbial Population Dynamics. Boca
Raton, Fla.: CRC Press, p. 55.
Baxerres, J.L., Haewsungcharern, A., and Gibert, H. (1977). Lebensm.-Wiss. Techno/.,
10, 191.
Bazin, M.J. (ed.) (1982). Microbial Population Dynanpcs. Boca Raton, Fla.: CRC Press.
Bentley, T.L., and Kincannon, D.F. (1976). Water Sewage Work, RIO.
Bergter, F. (1972). Wachstum von Mikroorganismen. Jena: G. Fischer Verlag.
Biggs, R.D. (1963). Canad. J. Chern. Eng., 60, 384.
Bischoff, K.B. (1966). Canad. J. Chern. Eng., 45, 281.
Bjorkman, U. (1987). Biotechnol. Bioeng., 29, 101 and 114.
Bienke, H. (1979). Adv. Biochem. Eng., 13, 122.
de Boks, P.A., and van Eybergen, G.C. (1981). Biotechnol. Lett., 3, 577.
Borzani, W., et al. (1976). Biotechnol. Bioeng., 18,623,885.
Brown, D.E. (1981). Instn. Chern. Eng., Ser. 64, N7.
Bruxelmane, M. (1983). Tech.l'1ngen., 2, A591O.
Bryant, J. (1977). Adv. Biochem. Eng., 5, 101.
Bryant, J., and Sadeghzadeh, N. (1979). Paper F3 presented at 3rd Eur. Conf. on
Mixing, Univ. of York, Great Britain. April 4-6.
Bull, D.N., and Young, M.D. (1981). Biotechnol. Bioeng., 23, 373.
Bu'Lock, J.D. (1965). The Biosynthesis of Natural Products. New York: McGraw-Hill,
Chap. 1.
Bu'Lock, J.D., et al. (1965). Canad. J. Microbiol., 11,765.
Bungay, H.R., III, and Bungay, M.L. (1968). Adv. Appl. Microbiol., 10,269.
Bungay, H.R., III, et al. (1969). Biotechnol. Bioeng., 11, 765.

Calderbank, P.H., and Moo-Young, M. (1959). Trans. Instn. Chern. Engrs., 37, 26.
Cameron, I.L., and Padilla, C.M. (eds.) (1966). Cell Synchrony. New York: Academic
Press.
Charaklis, W.G. (1981). Biotechnol. Bioeng., 23, 1923.
Charaklis, W.G., et al. (1982). Water Res., 0, 1.
Charles, M. (1978). Adv. Biochern. Eng., 8,1.
Charley, RC., et al. (1983). Biotechnol. Lett., 5, 169.
Chen, J.W., et al. (1965). Ind. Eng. Chern., Proc. Des. Dev., 4, 421.
Chi, C.T., and Howell, J.A. (1976). Biotechnol. Bioeng., 18, 63.
Chi, C.T., et al. (1974). Chern. Eng. Sci., 29, 207.
Constantinides, A., et al. (1981). Biotechnol. Bioeng., 23, 899.
Cooker, B., et al. (1983). Chern. Eng., Proc. Des. Dev., 19, 600.
Cooney, Ch.L. (1979). Proc. Biochern., 14, May, 31.
Cooney, Ch.L., and Acevedo, F. (1977). Biotechnol. Bioeng., 19, 1449.
Cooney, Ch.L., and Wang, DJ.C. (1976). Biotechnol. Bioeng., 18, 189.
Coughlin, RW., et al. (1975). Chern. Ing. Techn.,47, 111.
Coulman, G.A., et al. (1977). Appl. Environ. Microbiol., 34, 725.
Court, J.R., and Pirt, SJ. (1976). 5th Internat. Ferment. Symp., Berlin, 6.21.
Cysewski, C.R., and Wilke, Ch.R (1978). Biotechnol. Bioeng., 20,1421.
Daigger, G.T., and Grady, C.P.L. (1982). Biotechnol. Bioeng., 24,1427.
Dairaku, K., et al. (1981). Biotechnol. Bioeng., 23, 2069.
Danckwerts, P.V. (1958). Chern. Eng. Sci., 8, 93.
Dawson, P. (1972a). J. Appl. Chern. Biotechnol., 22, 79.
Dawson, P. (1972b). In Fermentation Technology Today (Terui G.,ed.) Proc. 4th Intern.
Ferment. Symp. Society of Ferment. Technol. Japan Osaka, p. 121.
Dawson, P. (1980a). In Bioconversion and Biochern. Engineering (Ghose, T.K., ed.) Proc.
2nd Symp., Indian Institute of Technology, New Delhi Vol. 2, p. 275.
Dawson, P. (1980b). Paper at 6th Intern. Ferment. Symp., London/Ontario July 20-25,
no. F-8.1.2.
Dawson, P. (1980c). Paper at 6th Intern. Ferment. Symp., London/Ontario July 20-25,
no. F-8.1.10.
Dean, A.C.R, et al. (1972). Environmental Control of Cell Synthesis and Function.
New York: Academic Press.
Deindoerfer, F.H., and Humphrey, A.E. (1959). Ind. Eng. Chern. 51,809.
Demain, A.L., et al. (1979). Proc. Syrnp. Soc. Gen. Microbiol., 29.
Dorawala, T.G., and Douglas, J.M. (1971). Amer. Inst. Chern. Eng. Journal AIChEJ,
17,974.
Dostalek, M., and Haggstrom, M. (1982). Biotechnol. Bioeng., 24, 2077.
Douglas, J.M. (1972). Process Dynamics and Control, Vol. 2. Englewood Cliffs, N.J.:
Prentice Hall.
Dunn, I.J., and Mor, J.R. (1975). Biotechnol. Bioeng., 17, 1805.
Eckenfelder, W.W. (1966). Industrial Water Pollution Control. New York: McGraw-
Hill, Chap. 13.
Edwards, V.H. et al. (1970). Biotechnol. Bioeng., 17,975.
Eggers, E., and Terlouw, T. (1979). Water Res., 13, 1077.
Eilers, H. (1941), Kolloid Z., 97, 313.
Einsele, A., and Finn, R.K. (1980). Ind. Eng. Chern., Proc. Des. Dev., 19,600.
Einstein, A. (1906). Ann. Physik., 19,289.
6.10 Final Note 399
Einstein, A. (1911). Ann. Physik., 34, 591.

Eirich, F., et al. (1936). Kolloid Z., 74, 276.
Erickson, L.E., et al. (1972). J. Appl. Chern. Biotechnol., 22,199.
Esener, A.A., et al. (1981a). Biotechnol. Lett., 3, 15.
Esener, A.A., et al. (1981b). Eur. J. Appl. Microb. Biotechnol., 13, 141.
Esener, A.A., et al. (1981c). Biotechnol. Bioeng., 23, 1851.
Fan, L.T., Erickson, L.E., Shah, P.S., and Tsai, B.I. (1970). Biotechnol. Bioeng., 12,
1019.
Fan, L.T., Tsai, B.I., and Erickson, L.E. (1971). Amer. Inst. Chern. Eng. Journal
AIChEJ, 17,689.
Fencl, Z. (1964). In Malek, I, et al. (eds.). Proceedings of Symposium on Continuous
Culture of Microorganisms. Prague: Csechoslovak Academy of Sciences, p. 23.
Fencl, Z., and Novak, M. (1969). Folia Microb., 14,314.
Fencl, Z., et al. (1969). In Perlman, D., (ed.). Fermentation Advances. New York:
Academic Press, p. 301.
Fencl, Z., et al. (1972). J. Appl.Chem. Biotechnol., 22, 405.
Fencl, Z., et al. (1978). In Proceedings of the 7th Symposium on Continuous Culture of
Microorganisms. Sikyta, B., et al. (eds.). Czechosl. Acad. of Sciences Prague: (1980)
p.49.
Fiechter, A. (1982). In Rehm, H.J., and Reed, G. (eds.). Biotechnoloby-A Compre-
hensive Treatise, Vol. 1. Deerfield Beach, Fla., and Basel, Verlag Chemie Weinheim,
Chap. 7.
Finn, R.K., and Fiechter, A. (1979). Symp. Soc. Gen. Microbiol., 20, 83.
Furusaki, S., and Miyauchi, T. (1977). J. Chern. Eng. (Jap.), 10 (3),247.
Gerhardt, P., and Gallup, D.M. (1963). J. Bacteriol., 86, 919.
Gerstenberg, H., et al. (1980). Chern. Ing. Techn., 52, 19.
Geurts, Th.G., et al. (1980). Biotechnol. Bioeng., 22, 2031.
Goma, G., et al. (1979). Biotechnol. Lett., 1,415.
Goto, S., et al. (1973). J. Ferment. Technol., 51, 582.
Greenshields, R.N., and Smith, E.L. (1974). Proc. Biochem., April, 11.
Grieves, R.B., et al. (1964). J. Appl. Chern., 14,478.
Gutke, R. (1980,1982). In UNEPjUNESCOjlCRO training course, Theoretical Basis
of Kinetics of Growth, Metabolism and Product Formation of Microorganisms. Jena:
Science, Academy of East Germany, ZIMET, Vol. 1, p. 112 (1980); pps. 39 and 58
(1982).
Gutke, R., and Knorre, W.A. (1980). Z. Allgem. Mikrobiol., 20 (7), 441.
Gutke, R., and Knorre, W.A. (1981). Biotechnol. Bioeng., 23, 2771.
Gutke, R., and Knorre, W.A. (1982). Biotechnol. Bioeng., 24, 2129.
Gutke, R., et al. (1980). Biotechnol. Lett., 2, 315.
Hamer, G. (1982). Biotechnol. Bioeng., 24, 51l.
Han, K., and Levenspiel, O. (1987). Biotechnol. Bioeng., in press.
Harremoes, P. (1978). In Mitchell, R., (ed.), 1. Wiley & Sons N.Y. Water Pollution
Microbiology, Vol. 2. Chap. 4.
Harris, N.P., and Hansford, G.S. (1976). Water Res., 10,935.
Harrison, D.E.F., and Topiwala, H.H. (1974). Adv. Biochem. Eng., 3, 167.
Heckershoff, H., and Wiesman, U. (1981). Chern. Ing. Techn., 53, 268.
Hegewald, E.M., et al. (1978). In Sikyta, B., et al. (eds.), Czechosl. Acad. of Sciences.
Proceedings of the 7th Symposium on Continuous Culture of Microorganisms. Prague:
(1980), p. 717.
Heijnen, J.J., et al. (1979). Biotechnol. Bioeng., 21, 2175.

Heijnen, J.J., and Roels, J.A. (1981). Biotechnol. Bioeng., 23, 739.
Herbert, D. (1961). In Proceedings of Symposium on Continuous Culture of Micro-
organisms. London: Elsworth R., ed. Society of Chemical Industry Monograph 12,
p.21.
Herbert, D. (1964). In Malek, I., et al. (eds.). Continuous Cultivation of Microorganisms.
Prague: Czechoslovak Academy of Science, p. 23.
Herbert, D., et al. (1956). J. Gen. Microbiol., 14,601.
Herzog, P., et al. (1983). Chern. Ing. Techn., 55, 566.
Holmes, D.B., et al. (1964). Chern. Eng. Sci., 19, 201.
Howell, J.A., and Atkinson, B. (1976). Biotechnol. Bioeng., 18, 15.
Hughmark, G. (1980). Ind. Eng. Chern., Proc. Des. Dev., 19,638.
Humphrey, A.E. (1978). Am. Chern. Soc. Symp., Ser. 72.
Humphrey, A.E. (1980). Adv. Biotechnol., 1,203.
Imanaka, T., et al. (1973). J. Ferment. Technol. (Jap.), 51, 558.
Irvine, RL., et al. (1980). J. Water Poll. Contr. Fed., 52,1997.
Jennings, P.A., et a!. (1976). Biotechnol. Bioeng., 18, 1249.
Joshi, J.B. (1980). Trans. Instn. Chern. Engrs., 58,155.
Joshi, J.B., et al. (1982). Chern. Eng. Sci., 37, 813.
Kan, J.K, and Shuler, M.L. (1976). Amer. Inst. Chern. Eng. Symp. Ser. 172,31.
Kan, J.K, and Shuler, M.L. (1978). Biotechnol. Bioeng., 20, 217.
Kargi, F., and Park, J.K (1982). J. Chern. Techn. Biotechnol., 32, 744.
Katinger, H. (1976). Paper presented at 5th Internat. Ferment. Symp., June 28-July 3
Berlin, No. 4.16.
Keller, R, and Dunn, 1.1. (1978a). J. Appl. Chern. Biotechnol., 28,508.
Keller, R, and Dunn, 1.1. (1978b). J. Appl. Chern. Biotechnol., 28, 784.
Khang, S.J., and Levenspiel, O. (1979). Chern. Eng. Sci., 31, 569.
Kincannon, D.F., and Sherrard, J.H. (1974). Water Sewage Work, R32.
Kipke, KD. (1984). In Process Variables in Biotechnology, Bioreactor Performance
working group chairman W. Crueger, Dechema Monograph, Chap. 20.
Kishimoto, M., et al. (1976). J. Ferment. Technol., 54, 891.
Kitai, A., et al. (1969). Biotechnol. Bioeng., 11,911.
Kjaergaard, L., and Jorgensen, B.B. (1979). Biotechnol. Bioeng., 21,147.
Klei, H.E., et al. (1975). J. Appl. Chern. Biotechnol., 25, 535.
Knorre, W.A. (1968). In Proceedings of 4th Symposium on Continuous Culture of
Microorganisms. Prague; Malek, J., et al. (eds.) (1969): Academia, Prague p. 225.
Knorre, W.A. (1980). In Beier, W., and Rosen, R (eds.). Biophysikalische Grundlagen
der Medizin. Stuttgart, New York: G. Fischer, p. 132.
Koga, S., and Humphrey, A.E. (1967). Biotechnol. Bioeng., 9, 375.
Kong, M.F., and Yang, P.Y. (1979). Biotechnol. Bioeng., 21,417.
Kornegay, B.H. (1969). In Proceedings of 24th Industrial Waste Conference. Purdue
Univ., Lafayette, Ind. p. 1398.
Kornegay, B.H. (1975). Mathematical Modelling of Water Pollution Control Bioprocess,
Keinath, T.M. and Wainielista, M. (eds.), Ann Arbor, Mich., Ann Arbor Science
Pub!. Inc.
Kornegay, B.H., and Andrews, J.F. (1968). J. Water Poll. Contr. Fed., 40, R460.
Kosaric, N., et al. (1980). Adv. Appl. Microbiol., 26, 147.
Kramers, H., et al. (1953). Chern. Eng. Sci., 2, 35.
Kuhn, H.J., et al. (1980). Eur. J. Appl. Microbiol. Biotechnol., 10, 303.
6.10 Final Note 401
Kiing, W., and Moser, A. (1986). Bioprocess Engng., 1, 23.

Kiing, W., and Moser, A. (1988). Biotechnol. Lett., in press.
Lacey, R.E. (1972). Chem. Eng., September 4, 56.
Laederach, H., et al. (1978). In Preprints 1st Europ. Congr. Biotechnology, Interlaken,
Switzerland, Sept. 25-29, Dechema, Frankfurt.
La Motta, E.J. (1976). Biotechnol. Bioeng., 18, 1359.
Lane, A.G. (1977). J. Appl. Chem. Biotechnol., 27, 165.
Lee, H.H., and Yan, B.D. (1981). Chem. Eng. Sci., 36, 483.
Lee, I.H., et al. (1976). Biotechnol. Bioeng., 18,513.
Lee, J.M., et al. (1983). Biotechnol. Bioeng., 25, 497.
Leegwater, M.P.M., et al. (1982). J. Chem. Technol. Biotechnol., 32, 92.
Levenspiel, O. (1979). The Chemical Reactor Omnibook. Corvallis, Ore.: OSU Book
Stores.
Levenspiel, O. (1980). The Monod Equation; A Revisit; Biotechnology and Bio-
engineering, 22, 1671-1687: John Wiley and Sons, Inc.
Liepe, F.,et al. (1978). In reprints, 1st European Congress on Biotechnology, Interlaken,
Switzerland, Part 1, p. 78.
Lilly, M.D., and Dunnill, P. (1971). Proc. Biochem., 6 (8), 29.
Lilly, M.D., and Dunnill, P. (1972a). Adv. Biochem. Eng., 3, 221.
Lilly, M.D., and Dunnill, P. (1972b). Biotechnol. Bioeng. Symp., 3, 221.
Lilly, M.D. (1982). J. Chem. Techn. Biotechnol., 32, 162.
Lippert, J., et al. (1983). Biotechnol. Bioeng., 25,437.
Luedeking, R., and Piret, E.L. (1959). Biotechnol. Bioeng., 1,431.
Luttman, R., et al. (1981). Eur. J. Appl. Microbiol. Biotechnol., 13,90,145.
Maiorella, B., et al. (1981). Adv. Biochem. Eng., 20, 43.
Malek, I., and Fencl, Z. (1966). Theoretical and Methodological Basis of Continuous
Culture of Microorganisms. New York: Academic Press.
Malek, I., and Ricica, J. (1969). Folia Microbiol., 14,254.
Margaritis, A., and Wilke, Ch.R. (1972). Dev., Ind. Microbiol., 13, 159.
Margaritis, A., and Wilke, Ch.R. (1978a). Biotechnol. Bioeng., 20, 709.
Margaritis, A., and Wilke, Ch.R. (1978b). Biotechnol. Bioeng., 20, 727.
Marsot, P., et al. (1981). Biotechnol. Lett., 3, 689.
Matsumura, M., et al. (1981). J. Ferment. Technol., 59,115.
Mattiasson, B. (1983). Trends in Biotechnology, 1, 16.
McGrath, M.J., and Yang, R.Y.K. (1975). Chem. Eng. J., 9, 187.
Melling, J. (1977). In Wiseman, A. (ed.). Topics in Enzyme and Fermentation Biotech-
nology, Vol. 1. Chichester: Ellis Horwood Ltd., p. 10.
Mersmann, A., et al. (1976). Int. Chem. Eng., 16,590.
Merz, A., and Vogg, H. (1978). Chem. Ing. Techn., 50,108.
Metcalf & Eddy Engineers. (1972, 1979). Waste Water Engineering Treatment, Disposal
and Reuse. New York: McGraw-Hill.
Metz, B. (1981). Biotechnol. Bioeng., 23, 149.
Metz, B., and Kossen, N.W.F. (1977). Biotechnol. Bioeng., 19,781.
Metz, B., et al. (1979). Adv. Biochem. Eng., 11, 103.
Metzner, A.B. (1957). Americ. Inst. Chern. Eng. Journal AIChEJ, 3, 3.
Middleton, lC. (1979). Paper A2 presented at 3rd Eur. Conf. on Mixing, York, Great
Britain.
Molin, G., et al. (1982). Eur. J. Appl. Microbiol. Biotechnol., 15,218.

Hermann.
Monod, J. (1950). Ann. Inst. Pasteur, 79, 390.
Mooney, U. (1951). J. Colloid. Sci., 6, 162.
Moreno, M., and Goma, G. (1979). Biotechnol. Lett., 1,483.
Moser, A. (1973). In Dellweg, H. (ed.). Proceedings of the 3rd Symposium on Technical
Microbiology. Inst. fUr Garungsgewerbe und Biotechnologie Berlin, p. 61.
Moser, A. (1977). Chem. Ing. Techn., 49,612.
Moser, A. (1980a). In Moo-Young, M., et al. (eds.). Proceedings of the 2nd Intern. Symp.
on Waste Treatment & Utilization, Vol. 2. Oxford: Pergamon Press, p. 177.
Moser, A. (1980b). In Ghose, T.K. (ed.). Proceedings of the 2nd International Symposium
on Bioconversion and Biochemical Engineering, Vol. 2. New Delhi: Indian Inst.
Technology, p. 253.
Moser, A. (1983a). Proc. Adv. Ferment., 83 (Suppl. Proc. Biochem.), 201.
Moser, A. (1983b). In Proceedings of the 33rd Canadian Chemical Engineering Conference,
Toronto, Canad. Soc. for Chern. Eng., Oct. 2-5. Vol. 2, p. 417.
Moser, A. (1984a). In Ghose, T.K. (ed.). Proceedings of the 7th International Biotech-
nology Symposium, Vol. 2. New Delhi: Indian Inst. Technology, p. 529.
Moser, A. (1984b). Acta Biotechnolog., 4,3.
Moser, A. (1987). In Crueger, W., et al. (eds.). Physical Aspects of Bioreactor Perfor-
mance report of working party Bioreactor Performance of Europ. Fed. Biotechnol.,
Dechema, Frankfurt, Chap. 4.
Moser, A. (1985a). In Rehm, H.J., and Reed, G. (eds.). Biotechnology-A comprehensive
Treatise, Vol. 2. Deerfield Beach, Fla., and Basel: Verlag Chemie Weinheim, Chaps. 15
and 16.
Moser, A. (1985b). In Moo-Young, M. (ed.-in-chief). Comprehensive Biotechnology,
Vol. 1. Oxford: Pergamon Press, Part 2, Chap. 4.
Moser, A. (1985c). Conservation & Recycling, Vol. 8. No. 1/2, 193-210, Pergamon
Press, Oxford.
Moser, A., Preselmayr, W., and Scherbaum, H. (1974). In Proceedings of the 4th
International Symposium on Yeasts, Part I, Klaushofer, H. and u., Sleytr (eds.), Univ.
of Bodenkultur, Vienna, p. 117.
Moser, F. (1977). Verfahrenstechnik, 11,670.
Moser, F. (1980). In Moser, F. (ed.). Grundlagen der Abwasserreinigung, Vol. 2. Munich:
Oldenburg, p. 431.
Moser, F., et al. (1977). Prog. Water Tech., 8, 235.
Moser, F., et al. (1979). Os terr. Abwasser Rund., 24, 83.
Moll, D.G. (1979). Ph.D. thesis, Massachusetts Institute of Technology, Cambridge,
Mass.
Mou, D.G., and Conney, Ch.L. (1983a). Biotechnol. Bioeng., 25, 225.
Mou, D.G., and Conney, Ch.L. (1983b). Biotechnol. Bioeng., 25, 257.
Mukataka, S. (1981). J. Ferment. Technol., 59, 303.
Mukataka, S., et al. (1980). J. Ferm. Technol., 58,155.
Mulcahy, L.T., and La Motta, E.J. (1978). Rep. no. 59-78-2, Environmental Engineering
Program, Dept. Civil Engng., University of Massachusetts, Amherst.
Nagai, S., et al. (1968). J. Gen. Appl. Microbiol., 14, 12l.
Nagel, O. et al. (1977). Chem. lng. Tech., 44, 367.
Nagel, 0., Hegner, B., and Kurten, H. (1978). Chem. Ing. Techn., 50, 934.
6.10 Final Note 403
Nagel, 0., Kurten, H., and Sinn, R. (1972). Chem. Ing. Techn.,44, 367.
Nestaas, E., and Wang, D.I.C. (1981). Biotechnol. Bioeng., 23, 2803.
Norwood, K.W., and Metzner, A.B. (1960). Americ. Inst. Chem. Eng. Journal, 6, 432.
Novick, A., and Szilard, L. (1950). Proc. Nat. Acad. Sci., Wash., 36, 708.
Oleszkiewicz, J. (1976). Environ. Protect. Eng., 2,85.
Oleszkiewicz, J. (1977). Prog. Water Tech., 9, 777.
Oosterhuis, N.M.G. (1984). Ph.D. thesis, Technical University, Delft, Netherlands.
Ottengraf, S.P.P. (1977). Biotechnol. Bioeng., 19, 141l.
Paca, J., and Gregr, V. (1976). Biotechnol. Bioeng., 18, 1075.
Paca, J., and Gregr, V. (1979). Enzyme Microb. Technol., 1, 100.
Park, Y., et al. (1984). Biotechnol. Bioeng., 26, 457, 468.
Parulekar, S.J., and Lim, H.C. (1985). Adv. Biochem. Eng., 32, 207.
Pickett, A.M. (1982). In Bazin, M. (ed.). Microbial Population Dynamics. Boca Raton,
Fla.: CRC Press, Chap. 4.
Pickett, A.M., et al. (1979a). Proc. Biochem., 13, November, 10.
Pickett, A.M., et al. (1979b). Blotechnol. Bioeng., 21,1043.
Pickett, A.M., et al. (1980). Biotechnol. Bioeng., 22,1213.
Pirt, S.J. (1974). J. Appl. Chem. Biotechnol., 24,415.
Pirt, S.J. (1975). Principles of Microbe and Cell Cultivation. Oxford: Blackwell.
Pirt, S.J., and Righelato, R.C. (1967). Appl. Microbiol., 15, 1284.
Pirt, SJ., et al. (1961). J. Gen. Microbiol., 25, 119.
Pirt, S.J., et al. (1983). J. Chem. Tech. Biotechnol., 33B, 35.
Powell, 0., and Lowe, J.R. (1964). In Malek, I., et al. (eds.). Continuous Cultivation of
Microorganisms. Prague: Czechoslovak Academy of Science, p. 45.
Powell, 0., et al. (1967). Microbial Physiology and Continuous Culture. London: Her
Majesty's Stationery Office.
Prochazka, J., and Landau, J. (1961). Colin. Czech. Chem. Commun., 26, 296l.
Pye, E.K., and Humphrey, A.E. (1979). Interim Report to U.S. Department of Energy,
p.79.
Rautenbach, R., and Albrecht, R. (1982). Chem. Ing. Techn., 54, 229.
Regan, D.L., et al. (1971). Biotechnol. Bioeng., 13,815.
Reisinger, C., et al. (1987). Paper at 5th International Conference on partition in
Aqueous Two-Phase Systems, Oxford August 23-28.
Remson, I., et al. (1971). Numerical Methods in Subsurface Hydrology. New York: John
Wiley.
Reuss, M., et al. (1919). Europ. J. Appl. Microbiol. Biotechnol., 8,169.
Reuss, M., and Buchholz, K. (1979). Biotechnol. Bioeng., 21, 206l.
Reuss, M., et al. (1980). Paper presented at 6th Internat. Ferment. Symp., London,
Ontario.
Reuss, M., et al. (1982). Chem. Eng., June, 233.
Reusser, F. (1961). Appl. Microbiol., 9, 36l.
Ricica, J. (1969a). In Perlman, D. (ed.). Fermentation Advances. New York: Academic
Press, p. 427.
Ricica, J. (1969b). Folia Microbiol., 14,322.
Ricica, J. (1973). Folia Microbiol., 18,418.
Ricica, J., and Necinova, S. (1967). Mitteil. d. Versuchsstation f Giirungsgewerbe Wien.
11/12,130.
Ringpfeil, M. (1980). Paper presented at 6th Internat. Ferm. Symp., London, Ontario.
Rippin, D.W.T. (1967). Ind. Eng. Chem. Fundam., 6 (4), 488.
Rittmann, B.E. (1982). Biotechnol. Bioeng., 24, 501 and 1341.

Rittmann, B.E., and McCarty, P.L. (1980). Biotechnol. Bioeng., 22, 2349, 2359.
Roe1s, J.A. (1980a). Biotechnol. Bioeng., 22, 23.
Roe1s, J.A. (1980b). Biotechnol. Bioeng., 22, 2457.
Roels, J.A. (1982). J. Chem. Techn. Biotechnol., 32, 59.
Roels, J.A. (1983). Energetics and Kinetics in Biotechnology. Amsterdam: Elsevier
Biomedical.
Roe1s, J.A., and van Suijdam, J.C. (1980). Biotechnol. Bioeng., 22,463.
Roels, J.A., et al. (1974). Biotechnol. Bioeng., 16, 181.
Rogers, P.L., et al. (1980). In Ghose, T.K. (ed.). Proceedings of the Second Symposium
on Bioconversion and Biochemical Engineering, Vol. 2. Indian Inst. Technology,
New Delhi p. 359.
Romanovsky, J., et al. (1974). Kinetische Modelle in der Biophysik. Jena: G. Fischer.
(German translation by W.A. Knorre and A. Knorre).
Russel, R.M., and Tanner, R.D. (1978). Ind. Eng. Chem. Proc. Des. Dev., 17, 157.
Russel, T.W.F., et al. (1974). Biotechnol. Bioeng., 16, 1261.
Ryu, D.D.Y., and Lee, B.K. (1975). Proc. Biochem., 10 JanuaryjFebruary, 15.
Ryu; Y.W., et al. (1982). Eur. J. Appl. Microbiol. Biotechnol., 15, 1.
Schneider, H., and Moser, A. (1987). Bioprocess Eng., 2, 129.
Schiigerl, K. (1977). Chem. Ing. Techn., 49, 605.
Schiigerl, K. (1982). Adv. Biochem. Eng., 22, 94.
Schiigerl, K. (1983). In Proceedings of NATO ASI, Mass Transfer with Chemical
Reactions, Izmir Turkey (1981) Vol. 1 (72),415.
Schultz, J.S., and Gerhardt, P. (1969). Bact. Rev., 33, 1.
Seipenbusch, R., and Blenke, H. (1980). Adv. Biochem. Eng., 15, 1.
Sheintuch, M. (1980). Biotechnol. Bioeng., 22, 2557.
Sherrard, J.H. (1977). J. Water Poll. Contr. Fed., 49, 1968.
Sherrard, J.H., and Lawrence, A.W. (1975). J. Water Poll. Contr. Fed., 47, 1848.
Sherrard, J.H. (1980). J. Chem. Techn. Biotechnol., 30,447.
Sherrard, J.H., and Schroeder, E.D. (1976). J. Water Poll. Contr. Fed., 48, 742.
Shieh, W.K. (1980a). Water Res., 14,695.
Shieh, W.K. (1980b). Biotechnol. Bioeng., 22, 667.
Shieh, W.K., et al. (1981). J. Water Poll. Contr. Fed., 53, 1574.
Shieh, W.K., et al. (1981). J. Water Poll. Contr. Fed., 53, 1574.
Shimmons, B.W., et al. (1976). Biotechnol. Bioeng., 18, 1793.
Shiotani, T., and Yamane, T. (1981). Eur. J. Appl. Microbiol. Biotechnol., 13,96.
Sinclair, C.C., and Brown, D.E. (1970). Biotechnol. Bioeng., 12, 1001.
Solomon, B.O., and Erickson, L.E. (1981). Proc. Biochem., February/March, 44.
Sortland, L., and Wilke, Ch.R. (1969). Biotechnol. Bioeng., 11, 805.
Stenberg, O. (1984). Ph.D. thesis, TU G6teborg/Sweden.
Stieber, R.W., and Gerhardt, P. (1979). Appl. Environ. Microbiol., 37, 487.
Stieber, R.W., and Gerhardt, P. (1981a). Biotechnol. Bioeng., 23,523.
Stieber, R.W., and Gerhardt, P. (1981b). Biotechnol. Bioeng., 23,535.
Stieber, R.W., et al. (1977). Appl. Environ. Microbiol., 34, 733.
Stiebitz, 0., et al. (1987). Knorre, W. (ed.). Proceedings of the UNESCO-training
course, Modern Biotechnology: Optimization of Fermentation Processes, Jena, East
Germany, Oct. 12-31.
Stucki, J.W. (1978). Prog. Biophys. Molec. BioI., 33, 99.
Sundstrom, D.W., et al. (1976). Biotechnol. Bioeng., 18, 1.
6.10 Final Note 405
Swartz, R.W. (1979). Ann. Rep. Ferm. Proc., 3, 75.

Tempest, D.W. (1970). In Norris, J.R., and Ribbons, D.W. (eds.). Methods in Micro-
biology, Vol. 2. Academic Press, London, N.Y. p. 259.
Toda, K., and Dunn, 1.1. (1982). Biotechnol. Bioeng., 24, 65l.
Topiwala, H.H. (1974). Biotechnol. Bioeng. Symp., 4, 68l.
Topiwala, H.H., and Hamer, G. (1971). Biotechnol. Bioeng., 13,919.
Trilli, A., et al. (1977). J. Appl. Chem. Biotechnol., 27, 219.
Tsai, B.I., Erickson, L.E., and Fan, L.T. (1969). Biotechnol. Bioeng., 11, 18l.
Tsai, B.I., Fan, L.T., Erickson, L.E., and Chen, M.S.K. (1971). J. Appl. Chem. Biotechnol.,
21,307.
Tyagi, R.D., and Ghose, T.K. (1980). Biotechnol. Bioeng., 22, 1907.
Vand, W. (1948). J. Phys. & Colloid. Chem., 52, 277.
van Dedem, G., and Moo-Young, M. (1973). Biotechnol. Bioeng., 17, 130l.
van Suijdam, J.C. (1987). In Crueger, W., et al. (eds.). Physical Aspects of Bioreactor
Performance report of working party Bioreactor Performance of Europ. Fed.,
Biotechnol. Frankfurt: Dechema. Chap. 6.
van Suijdam, J.e., and Metz, B. (1981). Biotechnol. Bioeng., 23, Ill.
van Suijdam, J.C., et al. (1982). Biotechnol. Bioeng., 24, 177.
van Suijdam, J.C. (1986). Paper at 14th Intern. Congr. of Microbiol., Manchester
Sept. 7-13.
van Suijdam, J.C., and Dusseljee, P.J.B. (1987). In Crueger, W., et al. (eds.). Physical
Aspects of Bioreactor Preformance report working party Bioreactor Performance,
Europ. Fed. Biotechnology, Chap. 6, Dechema, Frankfurt.
Warmoeskerken, M., and Smith, J.M. (1982). Paper Gl at 4th Europ. Conference on
Mixing Noordwijkerhout, Netherlands.
Wandrey, c., and Flasche1, E. (1979). Adv. Biochem. Eng., 12, 148.
Wandrey, c., Flasche1, E., and Schiigerl, K. (1979). Biotechnol. Bioeng., 21,1649.
Wittler, R. et al. (1983). Eur. J. Appl. Microbiol. Biotechnol. 18, 17.
Whitaker, A. (1980). Proc. Biochem., May, 10.
Wolfbauer, 0., Klettner, H., and Moser, F. (1978). Chem. Eng. Sci., 33, 953.
Woods, J.L., and O'Callaghan, J.R. (1975). Biotechnol. Bioeng., 17,779.
Wright, D.G., and Calam, C.T. (1968). Chem. Ind., 1274.
Wu, Y.c., et al. (1980). Biotechnol. Bioeng., 22, 2055.
Yamane, T., and Hirano, S. (1977). J. Ferment. Technol., 55,156.
Yamane, T., and Shimizu, S. (1982). Biotechnol. Bioeng., 24, 273l.
Yamane, T., and Shimizu, S. (1984). Adv. Biochem. Eng., 30,148.
Yang, Ren der, and Humphrey, A.E. (1975). Biotechnol. Bioeng., 17, 1211.
Yano, T., and Koga, S. (1973). J. Gen. Appl. Microbiol., 19,97.
Yoshida, F., et al. (1973). Biotechnol. Bioeng., 15,257.
Zeuthen, E., (ed.) (1964). Synchrony in Cell Division and Growth. New York: Interscience.
Ziegler, H. (1980). Biotechnol. Bioeng., 22, 1613.
Ziegler, H., et al. (1977). Biotechnol. Bioeng., 19,507.
Zlokarnik, M. (1978). Adv. Biochem. Eng., 8,133.
Zlokarnik, M. (1979). Adv. Biochem. Eng., 11, 157.
Zwietering, Th.N. (1959). Chem. Eng. Sci., 11, 1.
APPENDIX I
Fundamentals of Stoichiometry of
Complex Reaction Systems
Bioprocessing includes a large variety of metabolic reactions even on the

macroscopic level (cf. Fig. 2.16) and at the same time is carried out in different
modes of reactor operations. Therefore, complex reaction systems must be
treated stoichiometrically, which means that not only complex reactions
themselves must be considered but also complex reactor operations.
1.1 Stoichiometry of Complex Reactions

The stoichiometric equation in the case of complex reactions can be written
in the form
(U)
while for a simple reaction the following form is valid:
(1.2)
where i = number of reactions (1 :::;; i :::;; M)

j = number of components (1 :::;; j :::;; N)
Aj = components of reaction mixture
v = stoichiometric coefficients
The differential change of the number of moles n of component Aj due to the
reaction i is defined as
(1.3)
ei
where is the extent of reaction, defined as the change of number of moles
divided by the stoichiometric coefficient [mole].
In complex reactions, the singular reaction steps are interconnected, that
is, singular components participate in different reactions, with the conse-
quence that a part of the reactions is stoichiometrically dependent. Only the
independent reactions can be determined from the change in the number of
moles. The solution to the problem of stoichiometric dependence can be found
1.1 Stoichiometry of Complex Reactions 407
with the aid ofthe "matrix of the stoichiometrical coefficients" of the following
structure
~ Components 1 ~ j ~N
~
VI
VI (1.4)
where the row index is the number of reactions i and the column index is
the number of components j. Each row of the stoichiometric coefficient matrix
expresses the stoichiometry of a reaction in terms of the number of moles of
each compound converted per unit reaction rate.
The number of stoichiometrically independent reactions is given by the rank
of the matrix R p, which can be determined with e.g. the aid of the Gaussian
method of elimination. As a result, the stoichiometrical coefficients of Rv
linearly independent equations for the reaction system are necessary and
sufficient for, for example, calculation of the conversion of the key variables
and therefore also for all other components. Thus
(1.5)
Sometimes balancing is carried out without the formulation of reaction
equations. This situation, however, arises rarely when kinetics are of interest.
In this case an "element -species matrix" can be written on the basis of N
species (components) with k elementary balances
k'S1
..::.:
Components I
all ....
~ j ~N
...... alj
VI
VI a21 ... .... .. a2j
,.
(1.6)
en
.......
:::
E
Cl)
~l
...... ...... aij
Cl)
G:l
where row index = number of components, j

column index = number of chemical elements, i
i = number of elementary balances (1 ::;; i ::;; k)
408 I. Fundamentals of Stoichiometry of Complex Reaction Systems
a ij = number of atoms of atomic species i present in molecule

of componentj resp. aj' bj , cj ' dj , etc.)
The number of key variables R is determined again with the aid of the rank
of this matrix R. Thus (Schubert and Hofmann, 1975)
R = N - Rp (1.7)
Normally Rp = k in the case of N species and k elements.
Hence, only (N - k) net conversion rates can be chosen independently.
From this reasoning it becomes clear that the number of independent kinetic
equations to be postulated cannot be chosen at will-it is completely specified
by the number of elementary balances k and the number of components N in
the system (Roels, 1980). Which key variable or which kinetic equation is to
be chosen strongly depends on the application one has in mind.
As an example for the determination of the number of independent equa-
tions, the situation of an organism with balanced growth is considered. This
organism grows on one sole source of carbon and energy, a source that may
contain nitrogen. One sole source of nitrogen is supplied, and this source may
also contain carbon. One product is excreted; CO2 , H 2 0, and O 2 are the only
other components, to be exchanged with the environment.
In terms of formalism, the system of organisms will be considered to be a
given quantity of mass. The organism exchanges, macroscopically speaking,
an exact replica of itself with the environment; it is characterized by its gross
elemental composition formula Cal Hbl 0Cl N dl . The concept of the C-mole
of organism, that is, the amount containing 1 mole of carbon, is adopted (see
Sect. 2.2.3.2). Figure 6.53 gives a schematic representation of the system and
the possible flows to and from the system. Only the elements C, H, 0, and N
are considered, and indeed these elements comprise in most cases about 95%
of the cellular mass and the various other exchange flows. Equation 2.11 can
be directly applied to this specific case, with Vn; in the case of stirred tank
reactors being the rate of flow of components (Fj' C-mole/m 3 hr).
The element-species matrix that represents the elemental composition of
the flows of compounds can be written in this case as
k~1 C H N
FI : X bi ci dl
F2 : Sc a2 b2 c2 d2
F3 : P a3 b3 c3 d3
(1.8)
F4 : SN a4 b4 c4 d4
F5: 2 0 0 2 0
F6: CO 2 0 2 0
F7: H2 O 0 2 0
1.2 Example for Determination of Number of Linearly Independent Equations 409
In the present case there are seven flows, and Equ. 2.11 specifies four equations
between the flows represented in the matrix of Equ. 1.8. Hence, only three
flows are independent variables (cf. Sect. 1.2). Which kind offlows to be chosen
for measurement depends on the possibilities for experimental determination.
The knowledge, for example, of the respiratory quotient and the ratio of
oxygen consumption to substrate consumption allows direct estimation of the
biomass production rate and the product formation rate. This conclusion
from the application of balancing is of the greatest importance in situations
where process variables, for example, X, are very difficult to measure, which
is the case in penicillin fermentation (Mou and Cooney, 1983).
1.2 Example for Determination of Number of Linearly

Independent Equations ("Key Reactions")
Reaction scheme
(1.9a)
(1.9b)
Reaction steps
VA 'CA+ VBl'CB ~ VCl 'Cc (1.9c)
VCl 'Cc~ VA 'CA + VBl 'CB (I.9d)
VC2 Cc + VB2 CB ~ Vo' Co (1.ge)
Vo CD ~ VC2 Cc + VB2 CB (1.9f)
Process kinetics includes four rates of individual steps (r 1 to r 4) and the rate
equations for all compounds can be written on the basis ofrl to r4 as follows:
(1.1Oa)
(1.10b)
rC = V<;:l r l - VCl r 2 - VC2 ' r3 + VC2' r4 (1.10c)
ro = vO'r3 - vO'r4 (1.10d)
Thus, the matrix of stoichiometric coefficients is:
o
(1.11)
The rank of this matrix can be shown to be two by using the method of Gauss
410 I. Fundamentals of Stoichiometry of Complex Reaction Systems
elimination or by finding the order ofthe determinant, which is not zero. Thus,
for example, the second column is identical with the first when multiplied by
-1; the same with the fourth and the third columns. The remaining matrix is
then
0)
(~
o
-Va2 0
(1.12)
-VC2 0
VD 0
With Rp = 2, the number of key reaction or key variables is two, and these
can be chosen arbitrarily (e.g., A and D). Thus the remaining dependent rates
can be written on the base of r A and r D :
(1.13a)
(1.13b)
1.3 Stoichiometry of Complex Reactor Operation

The majority of stoichiometric considerations is restricted to closed reactor
operations, where div (cjv) = 0, so that according to Equ. 2.3b
dc
_J = +v .. r (1.14)
dt - J
A simple case appears also with continuous stirred tanks.
13.1 SEMIDISCONTINUOUS REACTOR

For semidiscontinuous processing, often used in fermentations for the produc-
tion of yeast biomass or secondary metabolites, the basic balance equation is
to be modified. Balancing in this case has to distinguish between components
that are already present in the reactor at the beginning (number of moles nj,o)
and components that are fed later on to the reactor (flux of moles liY>- The
stoichiometric balance in this case of semidiscontinuous reactor operation is
(1.15)
where nj = number of moles of component j

njO = temporal initial value of n
nY = spatial initial value of n
This operation contains the time t as an independent process variable, in
contrast to all batch reactor configurations.
1.3 Stoichiometry of Complex l{eactor Operation 411
1.3.2 NONSTATIONARY REACTOR OPERATION

Stoichiometry, for example, in the case of nonstationary modes of reactor
operation, needs another fundamental equation. For the CSTR the balance
of a component j is written as
f
dc c - cf! d~.
d/ + Y = vij de' (1.16)
which gives after integration

cj = cY + (cj,o - cY)e- t / f + I vij' ~*(t) (1.17)
i
where ~* is the modified extent of reaction.

Characteristically this equation contains again the time t as a process vari-
able. This term with the time t disappears and can be neglected in the case of
stationarity (if t T) and when cj,o = cy,
that is, in a batch reactor. Therefore
the stoichiometric balance equation of a discontinuous system can be formally
applied to a nonstationary continuous stirred tank. Similarly, the stoichiome-
tric equations of heterogeneous reactor systems with interfacial mass transfer
can be derived (Budde, Bulle, and Riickauf, 1981).
BIBLIOGRAPHY
Budde, K., Bulle, H., and Riickauf, H. (1981). Stochiometrie chemisch-technologischer
Prozesse. Berlin: Akademie Verlag.
Mou, D.G., and Cooney, Ch.L. (1983). Biotechnol. Bioeng., 25, 225, and 257.
Roels, J.A. (1980). Biotechnol. Bioeng., 22, 2457.
Schubert, E., and Hofmann, H. (1975). Chem. lng. Techn.,47, 191.
APPENDIX II
Computer Simulations*
This appendix contains a series of computer simulations that are thought to

represent the most significant basic kinetic models in bioprocessing. The
models are summarized in Table 11.1. The simulations in the figures also
contain the values of model parameters chosen for demonstration. Mainly
two different kinds of plots are presented, the first showing concentration/time
curves and the second the corresponding time curves of specific rates of
bioprocesses. The models are as follows:
1. Simple Monod-kinetics in batch operation:
Reaction scheme: S ~ X (II.1 )
Reactor balance equations (DCSTR):
rx = fl(S)' x (1I.2)
1
- rs = -' fl(S) . x (11.3)
YX!S
with kinetic equation (Monod type, cf. Equ. 5.38):
S
fl(S) = flmax Ks +S (11.4)
Model parameters; process variables: flmaX' K s, YX!S; Xo , So

2. Monod kinetics with lag time tL (cf. Sect. 5.3.3.1):
rx = fl(S, t) x (11.5)
1
-rs = -' fl(S,t)X (11.3)
YX!S
*This appendix has been added to the English edition of this book as a conse-
quence of a critical recommendation by 1.1. Dunn (Federal Technical University/
Lab of Technical Chern. Zurich). Most of the simulations were realized at Graz the
University of Technology, Austria, or at ZIMET, Central Institute of Microbiology
and Experimental Therapy, lena, East Germany (Fig. II.33 and II.34).
TABLE ILL Overview of simulated bioprocess kinetic models.
No. Bioreactor operation Process variables Kinetic model type Variation Equations Figures
DCSTR X,S Microbial growth: So ILl-II.4 II.!
Monad XO II.2
YXIS 11.3
/lmax II.4
Ks II.S
2 DCSTR X,S Monad with tL tL II.S, 1I.6 II.6
3 DCSTR X,S Monad with kd kd II.3, II.7 II.7, II.8
4 DCSTR X,S Monad with ms ms II.2, 1I.8 1I.9, IUO
5 DCSTR X,S Monad with k d , ms k d , ms II.7, II.S 11.11, 11.12
6 DCSTR X,S Monad with S inhibition K,s II.2, 1I.3, H.9 II'!3, II.!4
7 DCSTR X,S Monod with 2-S limitation KI2 II.10-II.13 11.15
8 DCSTR X,S Monad: diauxie Scrit 11.14-II.19 ILl6, ILl 7
K I2 (K R ) II.1S, II.! 9 ,....
~
9 DCSTR X,S,O Monad with O2 limitation kLia H.20- II.23 II.20, II.21
()
10 DCSTR X,S,P; multiinhibition model for microbial K,p, K,sp(K R ), II.24-I1.29 II.22-II.24, n.25-II.29 0
production (with repression and Kp, K,sx 3
'"d
multiinhibitions) I::
.....
<1l
11 SCSTR X,S,P Type 10 with S feed Fs 1I.24-11.30 II.30, 11.31 ...
t/)
12 DCSTR Sc, Sd' Sads Biosorption cJt 11.31-11.38 II.32
13 CSTR X,S Monod So 11.39, 11.40 and 11.33
a'
I::
Xo 1I.4 II.34 :
o
::l
en
;;;l
~
r-, -!:>-
<:> ,....
~ w
S
:::
~
::='"
414 II. Computer Simulations
x ... X (C) S + . . MY C) . . . p MODEL : '1'10 00 '
.
SO
11 . S GI L
0 0 0
~ a: 2 ) 1 . 0 GIL
"' 3) 2.0 GIL
0
a:
0
0 ...
0
0 0 0
0 0 0 3
~
3
0
... :::: "' :::: .,

0 0 0
M
Cl Cl
....
I
Ul X
"
d lil ill
>- N
L
.
2
ill " 0
'"
'"
.
0'
~
:5 ~
0
"!
. 00 1.00 q . DO 5.00 8.00 , .00 9.00 9 . 00 10 DO
FIG. 11.1 T (H)
K51
KO
.100
.000
I1YHAX~
KI
.500
.0
'I'XSl
KO
.SOO
.000
TLAG -
KP
. 000
.000
SCRIT -
K5C --
.0
.000
KI2
KZ .000
.0 PIMAX -
HO
.000
.000
YXP1
He
- .000
.000
YXP2 -
HS
.000
.000
KLA .0 O .000 YXO .000 YPO .000 YPS .000 XH .000 KISP - .0 KISX - .0 KIP - .0
X . .. x (C) S + . .MY 0 ... p MODEL: 'MONaD'

xo
0'
'"
OJ>
." ,
a:
1)
2)
3)
.01
.os GIL
.10
GIL
GIL
"a: i'l ....

""
n 0
" '"
~
!;:
~ "a: ~
W
r '"' Cl <.:>
>-
L
.
0
N
Ul
ill
x
ill
:il .
0
;;.
"a: ~
0 0 0
0 0 0
. 00 1.00 2 . 00 3 . 00 4 . 00 5 .00 6.00 , .00 9.00 9 . 00 1Q , OC
FIG. 11.2 T (H)
<51 '00 MYMAX- .soo YXSl 500 TLAG - .000 SCRIT - .0 KI2 .0 PIMAX - . .000 YXP1 - .000 YXP2 - .000
<D . 000 KI .0 KO 000 KP . 000 K5C .000 KZ . 000 HO .000 He .000 H5 .000
<LA .0 O .000 YXO .000 YPO . 000 YPS . 000 XH .000 KISP - .0 KISX - .0 KIP - .0
II. Computer Simulations 415
x ... X eJ . .. 5 + . .MY (') .. 1" MODEL: '''10 00'
.
YXS'
<>
11 .20
Ii! ~ 2 ) . 50
iii 3) . 70
fi! ~ '"
~
c <>
c <> ~
0
'"'
:::
<.:l
<>
'" :::
<.:l
.
m
I
....
>- ..;
. .
<>
Ul
<>
X
II!
I:
<> :i'
"' '"
j;: ("!
~
<> <> c<>

<> <>
. 00 1.00 2 . 00 3 . 00 4 . 00 5 . 00 6 .00 7 00 B 00 9 . 00 10 , 00
F IG. 11.3 T CH)

KS1 100 MYMAX- .500 YXS1 .500 TLAG - . 000 SCRIT - .0 KI2 .0 PIHAX - . 000 YXP1 - .000 YXP2 - .000
KO
KLA
.000
.0
KI
O
.0
.ODO
KO
YXO
. 0 00
. 000
KP
YPO
.000
.OOD
Kse
YPS -
.000
.000
KZ
XM
.000
. 000
HO
KISP -
. 000
0 ""
KISX -
.000
.0
MS
KIP -
.000
.0
x .. X eJ . .. 5 t- .. MY Cl .. . 1" MOOE"L : '''10''100'

"1YMAX
&~ 0-
Il.20 I / H
~ 2) .50 I / H
'" 31 .80 l/H
3
..
2
," ...
<> 0-
3
'"'" <>
'"
<>
"
0
I.... !l'
-:<.:l ..
0-
!<:
'"' 2
:::.
>- til <.:l
L ill 0-
:i'
'" ><
~ .
c
:i!
("! ~ ""!
g c g
c .,
. 00 1 , 00 2 . 00 3 , 00 4 . 00 5.ao 6 . 00 7 . 00 a 00 9 . 00 11) . 00
FIG. 11.4 T CHl

<S1 ' 00 MYMAX - 500 YXS, 500 TLAG - . 000 SCRIT- .0 KI2 .0 PIMAX- 000 YXPt .000 YXP2 - . 000
<D .000 KI .0 KD . 000 KP . 000 Kse .000 KZ .000 HO . 000 MP . 000 MS .000
KLA 0 O 000 YXO . 000 YPO . 000 YPS .000 XM 000 KISP - .0 KISX - .0 KIP .0
x . .. X ~. .. 5 + .. MY (') . . . p MODEL : . MONDO '

KS1
"..
1)
.01 GIL
"'" "
oD 2) .10 GIL
'" 3) .50 GIL
. 0
a: ..." 0
a:
"
0 "
0
0
0
0
~
3
!l
..; ........J il !l
I
" -'
'"
..... .....
>- N
..
0
(J)
." "x .. 0'
N
I:
~ ".. "'"
~ f': "
a:
"~- "" "" . 00 1.00 2 . 00 3 . 00 4 . 00 S . OO 6.00 7 . 00 8 . 00 9 . 00 10 . 00
FIG. II.S T (H)

1(51 '00 MYMAX - .500 YXSl . 500 TLAG - .000 SCRIT - .0 KI2 .0 PIMAX- .000 YXPl . 000 YXP2 - . 000
KD .000 KI .0 KO . 000 KP . 000 Kse .000 KZ .000 HO .000 MP . 000 MS . 000
KLA .0 O .000 YXO .000 YPO - .000 YPS .000 XM .000 KISP - .0 KISX - .0 KIP .0
X . .. X ~ . .. 5 + .. MY (') . .. p MODEL: . MONDO . fLAG '

flAG
1) 1.00 H
"
~ "
~
"
'"
N
2) 200 H
3) S.OO H
""! fjl ".

'"
~ "~ ""
0
..... " -'

I ....
..... "
:i:
(?
-'
.....
>- Ul <.::>
E co
..." "
-
n N
X
f': "
N "a:
~ ~
".
" "" "~

. 00 . 80 1.60 '2.40 3 . 20 4 . 00 4 ,ao 5 .60 6 . 40 7 . 0 8 . 00
FIG. 11.6 f ( Hl
KS' .003 MYMAX- .680 YXSl , Il03 TLAG - 1.000 SCRIT - .0 KI2 .0 PIMAX - . 000 YXPl - .000 YXP2 - 000
KD .000 KI .0 KO .000 KP .000 Kse . 000 KZ 000 MO .000 Me . 000 MS . 000
KLA .0 O .000 YXO .000 YPO . 000 YPS .000 XM . 000 KISP - .0 KI SX - .0 KIP - .0
x . .. X C!l S .,. .. '1'1' O.

'" ..,OOt'L ''''0'100 1(0 '
1(0
1 1 .00 1 l>-l
...
a
5; 21 .02 1 '>-l
11 .10 1 l>-l
~ ~
.0 0:
'"
<>
~ ~
.J
.....
(,)
." -:: (,)
~
"
(Jl
0
'"
X
.
a
N
a
"' "
oD
a a 3
"' '"
a
"" .00 2 . 00 ~ , I'}D 6.00 a . ~o 10 . 00 121)0 1A . 1)0 1S.t}Q 19 ':)0 20 ~o
FIG. II.7 T (Hl

K51 .100 MYMAX - 350 YXSl . aDo TLAG - .001) SCRIT - .0 K12 .0 PJMAX - .000 YXPl - . 000 YXP2 - 000
KD . 100 Kl .0 KO .000 KP .000 K5C .000 KZ .000 HO 000 MP .000 H5 000
KLA .0 O 000 YXO . 000 YPO .000 YPS 000 XM . 000 KISP - 0 KISX - .0 KIP
X ... X rI . .. S + . .MY C!l . . P MODEL, 'MONOD KD '

KD
.
1) .00 l/H
:;: 2) . 02 l/H
3) . 10 l/H
:il
M
""M
0
."
~ "
.,a
>-
L:
-
..."
1 2 3
""
.00 '2 .00 4 00 6 . 00 B . OC! 10 . 00 12 . 00 14 . 00 16 00 18 . 00 20 , 00
FIG. II.8 T [Hl

KG1 100 MYMAX- 350 YXS1 .aoo TLAG - 000 SCRJT- .0 K12 0 PIMAX- .000 YXP l . 000 YXP2 - .000
KO .100 Kl 0 KO .000 KP 000 K5C .000 KZ .000 MO 000 MP .000 HS .000
KLA 0 O .000 YXO .000 YPO .000 YPS .000 XM . 000 KISP - 0 KISX - .0 KIP 0
x ... X I!J . . .S . MY C) . . p MODEL : . MONaD . MS '
MS
1) .00 l/H
!:: li: 2) .03 l/H
.,;
3) . 10 l/H
li:
cO .
~
-X I
X 2
<>
.,~ <> X 3
....
.J
<::>
.
Q
..; ....
.J
<::>
<>
N
(f)
:J . x
<>
"
<> <>
N
M "' ~
Q
"'
3 ~1
<>
'" , :1)0 2.00 3:00 A. OO 5.00
~"
g , I)O 7 00
51 51
R OO
51
9 ,00
EJ
10 em
FIG. 11.9 (H l
K51 '00 MYMAX - 350 YXS1 dOD TLAG - 000 SCRIT - 0 KI2 0 PIMAX - 000 YXP1 000 YXP2 - . 000
KD .000 KI .a KO . 000 KP 000 KSC ODD KZ 000 MO 000 MP 000 M5 '00
<LA .0 o .000 YXO 000 yPO 000 yPS 000 XM 000 KISP - 0 KISX - 0 KIP a
X .. . X I!J . .. S + . .MY c.J . .. P MODEL: ' MONOD . MS '

MS
1).00 l/H
!:: 2) .03 l/H
3) .10 l/H
li:
M
Q
Q
M
0
.
<>
~
'!:
<>
-
m
>-
I:
:;,:
<>
~
3 2 1
<>
<>
. 00 " DO 2 . 00 3 . 00 4 . QO 5 . 00 6 00 7 00 e ' 0 9 . 00 10 00
FIG. II.10 T (H)
<51 '00 MYMAX - 350 YXS1 ADO TLAG - 000 SeRIT - 0 <12 .0 PIMAX - 000 iXP1 . 000 l'XP2 - . 000
<0 .000 KI 0 KO ODD KP 000 K5e DOD KZ 000 MO DOD MP .000 MS '00
<LA 0 O' 000 YXO ODD YPO 000 YP5 000 XH ODD KISP - .0 KISX - 0 KIP .0
x . .X C!J. .S + .MY (') . .P MODEL, 'MONDO KO MS '

KO MS
1) .02 .03 l/H
2) . 02 .10 l/H
..." "'" 3) .10 .03 l/H
; II)
4) .10 .10 l/H
""! "~
'" ~
~ .;
""
~
""
.... "..;'"
.J
r-------___-==2
<.) <.)
Ul
lil
x
."
N
3

~
..;
.."
""! "a;
g "'"
3 . 00 ILtlO S.on 6 . 00 1 00 8 . 00 9 . 00 10 . 00
FIG. 11.11 T (H)
KS' '00 MYI1A:( - . '350 YXS 1 ,a.oo TLAG - .000 SCRIT - .0 KI2 .0 PIHAX - .000 YXP1 .000 YXP2 - .000
.0 .000 KP . 000 KSC . 000 KZ .000 MO .000 MP . 000 MS .050
KO
<LA
. 050
.0
<I
0' .000
KO
YXO .000 YPO - . 000 YPS . 000 XM .000 KISP - .0 KISX - .0 KIP - .0
X ... X C'J . .S + .. MY (') .. . p MODEL, 'MONDO

KD MS
. . KD MS '
1) .02 .03 l/ H
2) .02 .10 l/H
3 ) .10 .03 l/H
4) .10 .10 l/H
"..;
'"
""M
a
."
~ "
"~
)-
:E
2
3
""
.00 1.00 2 . 00 3 . 00 . 00 5 .00 6 . 00 7 . 00 8 , 00 g.oo 10 00
FIG. 11.12 T (Hl

K51 . 100 MYHAX- .350 YXS1 .400 TLAG - .000 5CRIT - .0 KI2 .0 PIMAX - .000 YXP 1 - .000 YXP2 - .000
KD
KLA
.050
.0
KI
0'
.0
.000
KO
YXO
. 000
.000
KP
YPO -
.000
.000
KSC
YPS -
.000
.000
KZ
XM
.000
.000
MO
KI SP -
.000
.0 ""
KI SX -
.000
.0
MS
KIP -
.050
.0
x . .. X ~ .. . -' + .. M." C!l r> MODEL: ' S-INHIRITION '

KIS
1) s.n GIL
~ "~ 2) 10.0 GIL
- "-
31 SO.O GIL
<>
~ <>
cO <0
<> <>
<> <>
<Ii on
...J . :2 "'"
<>
~ ui l'
'" "'"
x
g
"
N
M
"'"
FIG. n .13
"" - "" "- -.--,
... 2 10 . ~, 6 ~~
" 00 '0
(H)
""
12 . QI) 14 . 00 ' 6 00 Hi 00 20 00
1(: 1 'O{'"l M"'1A!(- 3S0 f"<-" soc TLI\" 000 SCRI f- a Kf.-: .a PIMA>:: - .0 00 Y'':P1 000 YXP:? 000
<0 otJ'J
"o 2.0 1<0 "P 000
1)1"10 KSC 000 KI 001) MO 000 MP ,1)00 MS 000
KL 4 0 " >:0 01)0:' 000 y p .; . 000 KlSP
t;1)t) (DO . 1)01) XM .0 K : ;3X 0 KIP 0
x . . .x ~ . . .: + (') .. P MODEL: ' S-INHI BI TION'

KI
1) S.O S/L
"
N
2) 10.0 S/L
3 1 50 . 0 r;/ L
<>
"'
<>
<>
'"
.
0
<>
~ '"'
~
>-
:>::
-
~
3 2
"'"
"<>."
.00 ~ . ".I"; 4 . 00 G 00 ~ ro ' 0 . 1)1) ' 2 .~0 ,oa .!)/) 16 (II) 19 QI) 20 QO
FIG. 11.14 T [H)
10'9' '00 M:'MA::Z- 350 '(7..;3+ - ,SOD ruv:; 000 ~~ CRTT - 0 .I~ 0 PI/"IA)( - 000 y'.;r> 1 0(1[' YXP.~ rr;f)
KD 000 KI 2 .0 KO 000 KP ODD KSC oeD "Z 000 MO 000 MP 000 MS 000
~LA 0 o ODD YXO - ODD YPO 000 ypS oeo XM OCr) KISP 0 KISX 0 :TP .0
x ... X ~ .. .5 +. .MY (!) . . . 52 MODEL, '2 -S-MONOO '

KI2
.." 1.0 GIL
1)
e- "
ll) "
~
2) 10.0 GIL
"
~ ~ .
" ""'
"
"" "a: "" "on
0 "
..J "
.... "' ~ 1& "~ ....
..J "
Q
-:! "x
'"
(Il
Ii!
>-
I: W " "
"
~
(I] -
.
., ,-: c
~
c
'"
"'" !!' ." ~
"" "" "" "" . 00 .80 1.60 2 . 40 4 .00 4.80 5 .60 8 .00
FIG. 11.15 T (H )
KS 1 .005 HYMAX- .670 YXS 1 - ,480 TLAG - .000 SCRIT- .0 KI2 1.0 PIMI\X- .000 YXP1 - .000 VXP2 - .000
.0
-
KO . 000 KI KO .000 KP . 000 Kse .000 KZ .000 Me .000 MP . 000 MS .000
KLA .0 O' .000 YXO . 000 YPO .000 YPs - .000 XM .000 KISP - .0 KISX - .0 KIP .0
X .. .X O . .5 1- . . MY C) . . . p MODEL, 'DIAUXTE '

SCRIT
1) .OS Gil
:z 0>
'" $ 2) S.OO Gil
.
" g .
"
so: '"
'"
.;
..."
:! ~
:! .,"'" :! ~
"
Q Q
(l (J'1 '><
;: '"'" ;:<
"'
!l' '"'" !l'
.'" "'" ~
"" "" "" .00 .u. . C1C1 9 .00 12.00 16 . 00 20 , 00 24 , 00 <9 . 00 32. 00 36 . 00 11.0 . 00
FIG. 11.16 (H '
KS' 035 MYMAX- ,450 YXS1 '50 TLAG - 1.500 seRIT- a KI2 10.0 PIMAX - .225 YXP1 . 380 YXP2 - .590
KO .000 K! 8.0 KO 000 KP .050 Kse . 000 KZ .000 MO .000 MP .000 MS .000
KLA .0 O' .000 YXO 000 YPO .000 YPS .250 XM .000 KISP - .0 KISX - .0 KIP 8.0
X ... X 1!l ... S + .. MY (!) . . . P MODEL, 'DIAUXIE'

SCRIT
1) . 05 GIL
2) 5.00 GIL
Ii:
<>
~
~
M
I
....
~
1:

<>
N
<>
0
. 00 4 . 00 8 . 00 12 . 00 16 . 00 20 . 00 24 . 00 28 00 321)1) 36.00 tl.0 . 00

FIG. 11.17 ( HI
KS' .035 MYMAX - , 450 YXS1 .150 TLAG - 1.1500 SCRIT - .0 KI2 10 . 0 PIMAX- . 225 iXP1 . 380 (XP2 - .1590
KD .000 KI 8.0 KO . 000 KP .050 Kse . 000 KZ .000 Me .000 MP 000 "5 .000
KLA .0 o .000 YXO .000 YPO . 000 YPS - .250 XH .000 KISP - 0 K!SX - .0 KIP 8. 0
X . . X I!l. .. S + . .MY (!) . . . P MODEL, 'DIAUXIE'

KI2
1) .1 GIL
<> <>
2) .5 GIL
m <> ~ 3) 1.0 GIL
to :!. 4 ) 2.0 GIL
.
<>
N
<>
<>
~
<> <> <>
<>
to
<>
~
"
<> <> ~
-:! '" .J
<:)
....l) <>
ai -:!
l)
a. 1Il )(
:<: <> ~
<>
.;
.,
<> <>
<>
.;
'Ii
.
<> <>
<> ~
"
0 0 <>
0 0 <>
. 00 11 . 00 8 00 1~ . OO 16 . 00 20 . 00 24 . 00 28 00 32 . 00 36 , 00 110 , 00
FIG. IU 8 T (HI
K5 1 . 035 MYMAX- . .:1.50 YXSl 'SO TLAG - 1.500 SCRIT - , K12 2 .0 PlMAX- 225 YXP1 - .380 VXP2 - 590
KO 000 KI 10.0 KO 000 KP 050 Kse 000 KZ 000 HO . 000 HP 000 HS 000
KLA .0 0' .000 YXO 000 YPO 000 YPS .250 XH 000 KISP - 0 KISX - 0 KIP 20 . 0
x ... X [!) .. .S +. .MY C) . . . I" MODEL. 'DIAUX1E'

1(12
11 .1 GIL
21 .5 GIL
1 .0 GIL
0
3)
'" ~ l 2.0 GIL
0
0
#
:<:
..;
~
:;'
>- ..;
:E
.
0
0
"!
. 00 0. . 00 a 00 12.00 16 . 00 20 . QO 24 . 0{] 29 00 32 . 0D 36.QO 40 . 00
F IG. 11.19 T (HI

KS ' .035 MYHAX- ,450 YXSl 'SO RAG - 1 sao SCRIT - , KI2 2.0 PIMAX - . 225 YXP, - .380 YXP2 - S90
KD
KLA
. 000
.0
KI
O'
10.0
. 000
KO
YXO
OOD
. 000
KP
YPO -
.050
. 000
KSC
YPS
.000
. 250
KZ
XH
.000
. 000
MO
KISP -
. 000
.D ""
KISX -
. 000
.0
HS
KIP
.000
20 . 0
X ... X [!) .. .S + .. MY C) . . . 1" MODEL. . OTR- LIMIT ATION'
, 0 0
'" " '"
,... 0 ,
0
"
"
il 0
">'
0
"
S'
!l' ::: :::
0 0
()
"a: ()
"'
,,J
()
111
CI'
<0
0 '" :<:
0
'"
:1. "co .,
0
"
C- o
co .
0
0 co 0
co
" 0
.00 a.Clo '6 . 00 24 . 00 32 . 00 40 , 00 48 , 00 58 . 00 64 .00 n .oo ao .oo

FIG. 11.20 T (HI
K5 1 000 MYMAX- .10tl. YXSl 200 TLAG - . 000 SCRIT - .D KI2 .D PlHAX- . 000 YXPl . 000 YXP2 - .000
KD 000 KI .0 KO .00 1 KP .000 KSC ' 00 KZ . 000 MO .000 MP .000 MS . 000
<LA 10.0 O' .008 YXO . 880 YPO .000 YPS .aoo XH . 000 KISP - .0 KISX - .0 KIP .0
x . ..X ICI ... S + .. MY (') ... P MODE L, 'O TR-LIMITAT ION '
.
'< LA
0
11 10 1/H
21 30 1/H
31 250 1/H
"
0
0
'"
,I
>-
L
.
0
"
'"
0.
. 00 8 . 00 16 IJO 24 00 32 . 1')0 40 ")0 .a8 . ")0 56 00 7} ,10 91 'l'l

"" 00
FIG. 11.21 rf-jl
KS' 000 I1),MAX - 'a" 1'X5 1 200 TLAC - 000 seQ! r- a <12 0 P I I1AX - 000 iXP1 000 n: P 2 - IJ~ Q
<a 000 <I .D <0 . 001 <P . ODa <so ' aD <l OOD "0 ODD
"" 000
a
M5 'J(; 'J
a
<LA 10 . 0 0' 008 YXO AAO YPO ODD 'P5 . 000
'" ODD '< ISP - 0 '( 15)( <1 P
X ... X ICI . . . S + .. MY (') ... P MODEL, ' MUL TI-I NHI BIT ION '
'<IP
11 . 2 GI L
0
'"
0
0 '"
c
21 2 . 0 GIL
~ ;; 3 1 50.0 GI L
.
0 0
'"~
0
'"
'" ~
'"
'"
....
0
'"
1':
~
~
~ '"0
el el e
a. (fl x
~ '"0
~
n
"'" '"
Q)
.
'" <-
'"
c
'" "'"
. 00 , 00 2 00 3 00 00 5 . 00 6 . 00 7 00 8 00 9 00 10 ,00
FIG. 11.22 T (f-jl " 0'

KS ' - 2.000 '1'1'111\ >(- . 020 fX S' 7SD SCfED- 0 ~EED - 0000 VD 30 . 0 P!MAX - . 003 ,(XP l .000 YXP2 DDD
<0 . 00 1 <I .0 <0 , 000 <P 500 <50 . 000 KZ . 000 MD . 000 MP .000 M5 . 0 10
<LA - 500 . 0 o 000 YXO . 250 ,"" - 0 . 250 iDS "D XM . 000 K!SP - 30,0 KISX - 80 . 0 KIP 20 . 0
x . .. X [!J . S + .. ""Y C). .. P MODEL, . MUL Tr - INHIBIT ION '
I(IP
....;" .."..; 1)
2) 2.0 GIL
.2 GIL
3) 500 GIL
.'" .
"
'"
"
"" ""
0 '" N
0 '"
""!
::r
-... ::r
-...
l'l l'l
qp - jJ - 3
P
"
<D
"
II!
.
" ;:
"" "~
. 00 1.00 2 . 00 3 . 00 4: . 00 $ . 00 6.00 7 . 00 B. OO g . OO l lC . OO
FIG. II.23 T CHl _10 1

K51
KD
- 2.000
.001
MYMAX~
KI
.020
.0
YXSl .750
.000
SFEED - .0 FEED - . 0000 VO 30.0 PIHAX- .003
,.,
YXPl - .000 YXP2 - .000
-
KO KP .500 KSC .000 KZ .000 NO . 000 .000 MS 010
KLA - 500.0 O' .006 1xa 1.250 YPO 6.250 YPS .710 XM . 000 KISP - 30.0 KISX - 80.0 KIP 20.0
X. . .X [:J . S + .. MY C) . .. P
.,...
!Ii "" "
D ""
~ ; ~
a
!fI "" ""
"' ""
~
'"
~
N
~ ""
" 0 ""
0
""0
<:> N " ~
"
<:>
'"
..J
;: -... "
.. C) "m
.,
"'
..J
..... "
"<D C) "..q .
2: ! -:!
(!)
qp : D. "" jJ "'" "cr; X

"<>
"' Ul
~
"" g
III '"" :<l
..; iii
..,., ;-
"
" f.l ""
N
g
is
"" "" "~
. 00 4 .00 8 . 00 12 . 00 16 . 00
,----,-----r
2[1 DO 2rt . OO <9 . 00 32 .00 36 . 00 Llei . OO
F IG. 11.24 h -10'

KS 1
KO
2.000
.001
MVMAX-
KI
.020
.0 Ka
YXS1 , 75 0
.000
TLAG -
KP
.000
.500
SCR IT-
KSC
.0
.000
KI2
KZ
.00
.000
PIMAX- .003 YXP1 - .000 YXF2 - . 000
.010
-
MO .000 MP .000 MS
KLA - 500 . 0 a . a06 yxa 1.250 ypa 6.250 YPS - .7 10 XM .000 KISP - 30 . 0 KISX - BO.O KI P .S
g KP KI P KISP KISX
( I) 0.002 1.0
4 [I Il 0.020 1.0
iii ( III ) 0.002 10.0
[ IV ) 0 . 020 10 .0
'""!
x
<{
E N
~
a..
....
.....
a..
~
i<l
"" ,~~ :'0 , >0 . 30 . -0 ,so . &0 . 70 ,aD . '10 1.')0
FIG.II.25 '1Y/'1Y'1AX
xo 10.') SO - 100 . 0 Y)(/S . 1.50 (P I S .gOO t1y!'1AX - .0920 PlMAX - . 0050 <5 . i SO
'is - . 1) 111') <0 - . !JOi O
." KP KIP KISP KI5X

(I) 0 .50 200.0 1.0 1000 .0
200.0 10 . 0 4 1000.0
'" 0.50 20 1000.0
0 ,50 200.0 1000.0 1000.0
X '"
~,
is
x
<{
:r:
..... ~
a..
....
.....
a..
!ZI
!"
2
""
, DO . '0 . 20 , 3D . 40 . 50 . 60 . 70 .aD . 90 1,00
FIG. 11.26 MYIMYMAX

xo
MS
10.0
- .0100
50
KD
- 100 . 0
- . 00 10
YX/S - . 750 VP/5 - . 7 10 MYMAX - .0200 PIMAX - .0030 K5 - 2 . 000
.
0
KP KIP KISP KISX
(I ) 0.05 200.0 80.0 1000.0
(II) 0 . 50 200.0 80.0 1000.0
( II II 5.00 200.0 80.0 1000.0
( IV) 50.00 200 . 0 80.0 1000.0
~
x
is
x
-0:
5 "
"!
-
a.
.....
a.
II!
4
III
!:>
0
0
. 00 . '0 .20 . 30 .<0 . 50 . 60 , 70 .ao . 90 1.00
F IG. II .27 MY IMYMAX

xo 10.0 50 - 100 . 0 YX/S - . 750 yP/S .710 MYMAX - .0200 fllHAX - .0030 K5 - 2.000
HS - . 0100 KD - .0010
.0
KP KIP KISP KISX
( I) 0.50 0.2 80.0 1000 . 0
( II) 0.50 2.0 80.0 1000 . 0
~ (III ) 0.50 20.0 80.0 1000 . 0
( IV) 0.50 200.0 80.0 1000.0
~
X
x
-0:
5
-
~
a.
.....
a.
~
III
!:>
0
0
. 00 , '0 . 20 . 30 ,< 0 . 50 . 60 .7 0 .ao . 90 1.00
F IG. n .28 MY/MYMAX

xo 10 . 0 SO - 100 . 0 YX/S .750 YP/S . 710 MYMAX - .0200 PIHAX - .0030 K5 - 2.000
M5 - . 0100 KD - . 0010
." KP KIP KISP KISX

(I) 0.5 80.0 80.0 100.0
(II) 0.5 80.0 80.0 150.0
;; (III) 0.5 80.0 80 . 0 200.0
( IV) 0 .5 80.0 80.0 500.0
?!
X
3
ill
-
x

I: !;l
-
n.....
n.
~
!:'
""
. 00 . '0 .20 . 30 . <0 . SO .60 . 70 .80 .90 1.00
FIG. 11.29 MY/MYMAX

XO
MS
10 . 0
- .0100
so
KD
- 100.0
- .0010
YX/5 - .750 VP/5 - .710 MYMAX - .0200 PIMA)( - .0030 K5 - 2.000
x ... X [!J . . . S + .. MY (!) .. .P
F/VO
1, .0005 l/H 3
"" g "" l/H
2.') .001
re ~ III 3, . l/H
"" "" g .3
Cl; ~ ~
"" 0 "c c
c
~ ;i g 2
.J " g
(-, :4" "" :::!m
<.:> '"
"
-
.J
Z ....
U
a. "" X "
c
c
"~
~
r.n '"
g "" ""
.; :4
""
" .
""""
's......
............. EJ.... _ - --+J 3
"" "" ""! ~~
. 00 4 00 B. OO 12 . 00 16 . 00 :10 00 :24 . 00 28 00 32.0D 36 . 00 110 . 00
FIG.II.30 h lO'
K51 2 .000 MYMAX- .050 YX5 1 750 SFEED 500 .0 FEE:'"D - .O O2C "0 35.0 PIMAX- .003 YXP1 - .000 YXP2 .000
KD OD1 KI .D KO .000 KP 500 KSC .000 KZ .000 MO .000 MP .000 MS .010
KLA 500 .0 D . 006 YXO 1. '2 50 H'O - 6.:.!SO YPS .7 10 XM . 000 KISP - 30 . 0 KISX - 50 .0 KIP 20.0
FIVO
1, .0005 1/H
il> ~. 2, . 001 1/H
3 : . 002 1/H
~ ~
'"
6 ""' ~
Cl "",'
<>
~ ."
I
.... ~
qp
lit "
N
/J
~
~
"
~ "
d
"" g
. 00 4 . QO e.e'J 12 , QO IS .ao
;
24 . 00
!C
28 . 0 0 32.00
;
~.OD
; !~
"0 ,00
FIG. II.3\ h .10'

KSI . 2 . 000 MYMAX- .050 YXS1 . 750 SFEED- SOO.O FEFD - .0020 vo 35 .0 PIMAX- .003 YXP1 - .000 YXP2 - 000
KD .001 KI .D KO . 000 KP SOD KSC .000 I'.Z .000 HO .000 HP .000 HS . 010
KLA - 500.0 O' . 006 YXO 1 . 250 1PO - 6 . 250 YPS .710 X" .000 KISF' - 30 . 0 KISX - 50 . 0 KIP 20 . 0
810 sopp rIO"I
S'
.:!
u
(Jl
~ sads
SAOS. MAX
{ .
c>
<>
'" .00 '0 20 . 50 .60 .70 . 90 . 90 1 ,00
FIG. 1132 T (H l
- 23 . 10 "I - 3 0 .00 01 - .5000 SRMAX - . 1730 - 1.210 K - 1 . 160

----- - - --------------- -------- -- ~~;l

's So ' (g.' -')
------ -- ---- ---------- ----- -----r-------

.8
------ ---- ------ --------- ------. ~------

.6
.5
_ - - - ;.6-_ _ _ _ _ ~
., ~,
~1
o ~-=-=-=-=-=-=--'::-=--="-::::~'""2_=:::_==:=-.::..:-::.;--=-~-~-..:.-.::..-.::..--::.:-:..:-:.:-..:.-l:::===::;j
o .5 D
FIG. II.33
X
(go(')
.5
o UL_ _ _ _ _ _ _ _ _ _ _ __ _ ~ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _u
o 1- l00h
FIG. II.34
with
(II.6)
Model parameters: Jlmax, Ks, Yx1s , t L.

3. Monod kinetics with death rate kd (cf. Sect. 5.3.4.1):
rx = Jl(s)x - kd . x (11.7)
1
-rs = - ' Jl(s)x (II.3)
Yxis
with Equ. 11.4 for Jl(s) and Jlmax, Ks, Yxls , k d

4. Monod kinetics with endogeneous metabolism (ms) (cf. Sect. 5.3.4.2):
rX = Jl(s)x (11.2)
1
-rs = -Jl(s)x - ms' x (II.8)
Yxis
with Equ. 11.4 for Jl(s) and Jlmax, Ks, Yxls , ms; xo, So
5. Monod kinetics with kd and m. as a combination of Equs. II.7 and II.8.
6. Monod kinetics with S inhibition (cf. Sect. 5.3.5.1). Thus, taking Equs. II.2
and II.3 with Equ. 5.88:
1
(11.9)
Jl(s) = Jlmax 1 + Ks/s + s/K,s
and K,s being the key parameter.
7. Monod kinetics with double substrate limitation (cf. Sect. 5.5):
rx = Jl(S1' t)x + Jl(S2)X' f (11.10)
1
-rS1 = - - ' Jl(S1, t)x (II. 11)
YX1S1
1
- rS2 = - - ' Jl(S2)X . f (II. 12)
YX1S2
with Equ. II.6 for Jl(S1' t) and the following expression for the term f (cf.
Equ. 5.106 or Equ. 5.152):
(II. 13)
with KR being the key parameter for repression and s1,crit representing the
substrate concentration, where both substrates are utilized simultane-
ously, thus losing the diauxic behavior.
8. Diauxic growth as an analogy to case 7:
scheme: s.------"X
1------. P = S2 --+ X
rX = fl(Sl,t)X + fl(p)x'f1'f2 (II.14)

1 1
-rS1 = - - ' fl(Sl' t)x + - ' qp(s)' X (11.15)
YX1S1 YPls
(11.16)
with kinetic equations
S2
flS2 = J1m.x, 2 K
S2
+ S2 (S2 == p) (II.I8)
(II.I9)
and Equ. II.I3 for the term fl' f2 = f This simplest model of diauxie thus
contains 13 parameters: flm.x,l, J1max,2, K S1 ' Kp(== K S2 )' YXIS1 ' YxIP ' YPIS1 '
YXIS2 ' t Ll , K ls , KIP, K R (== K 12 ), and sl,crit
9. Monod with O 2 limitation (cf. Sect. 5.5.3):
rX = fl(S,O)X (II.20)
1
-rs =-'J1(s,o)x (II.21)
Yxis
(11.22)
with Equ. 5.169 for J1(s, 0):
fl(S,O) = J1m.x Ks
S
+S Ko
+ (II.23)
The parameters are flm.x, K s , Ko, Yx1s , Yx1o , kLl a, and 0*.
10. "Multiinhibition model" for the quantification of secondary metabolite
productions (cr. Sect. 5.4):
scheme: S +0 ~X +p
rX = fl(S,O)X - kd . x (II.24)
1 1
-rs = - ' fl(S,O)X + ms' x + - ' qp(s,o)x (11.25)
YxlS YPls
rp = qp(s,o)x (11.26)
ro = kLl a(or - ~) + rno' x - _1_ Jl (s, o)x - _1_ qp (s, o)x (11.27)
Yxlo lPlo
with the kinetic expressions for Jl and qp

s 0
Jl(s, 0) = Jlmax s (1 + s/KISX ) + K S 'x/(1 + p/KIP ) Ko +0 (11.28)
s
qp(s,o) = qP,max Kp + s(1 + s/KISP ) (11.29)
Combining Equs. 11.4 and 11.29 yielded Equ. 5.138 (K ISP == KR == KI!)
according to an approach of Bajpaj and Reuss (1981), while the full model
of "multiinhibitions" according to Moser and Schneider (1988) contains
four constants Kp, KIP, K isp (== K R), and K lsx ' The main aim of this
complex model was to achieve plots of increasing qp at decreasing Jl, as
illustrated in Figs. 11.23 and 11.24. The corresponding successful plots of
concentrations are shown in Figs. 11.22 and 11.24.
The behavior of the product formation model including such multi-
inhibition kinetics (with repression), basically represented in Fig. 5.47, is
shown in sensitivity analysis in the qp/Jl plots of Fig. 11.25 through 11.29,
with variations in all four parameters Kp, KIP' K ISP ' and K ISX '
11. Evaluation of substrate feeding strategy for a fed-batch culture with opti-
mal production of secondary metabolites-case study. On the basis of the
mathematical model (Equs. 11.24-11.29) and experimental estimation of
model parameters, an optimal S feed can be found by simulations adding
the expression Fs(t) for S feed to Equ. 11.25:
F
Fs(t) = -'SF (11.30)
Vo
where SF is the substrate concentration [g ,1-1 ] in the feedstream and Fs

is the substrate feed rate [gil' h], while F is the volumetric flow rate
[1. h -1] and Vo [1] is the initial volume. Figure 11.30 illustrates the concen-
tration/time curves for substrate, biomass, and product at varied S feed
F/Vo at constant SF' The corresponding time curves of specific rates Jl
in respect to qp show that choice of the right F/Vo allows the value for Jl
to be maintained at a minimum while qp is kept constant at a high level
(Fig. 11.31).
12. Biosorption model (cf. Sect. 5.3.9). A simple reaction scheme (see Fig. 5.40)
is written as a sequence of elimination and degradation
SL elim) Ss + 0 degrad) p (11.31)
The balance equations in case of a DCSTR are (cf. Equs. 5.112-5.115):
(II.32)
(II.33)
and the biosorption rate rads is

(II.34)
According to the literature (Theophilou et ai., 1979; see Sect. 5.3.9), a
sorption capacity ~ads is defined (see Equ. 5.115) and used in an analogy
to Langmuir adsorption (see Equ. 5.114). The kinetic constants were
elaborated by these authors to be load dependent:
k e = (0.58' fi)x (11.35)
kdegr = (1.86' L - 0.37)x for L < 0.9 h- 1 (11.36a)
kdegr = (-0.23' L + 1.53)x for L > 0.9 (1I.36b)
and
Se. = 41 fi + 12 [mg COD '1- 1 ] (II.37)
"Sludge loading" s (= adsorbed concentration Sads) was found to be L
dependent
s= 36.5' L + 22 [mg COD '1- 1 ] (11.38)
and sads,max == smax = 173 mg .1- 1 in this situation. Figure 11.32 represents a
typical plot of the time curves of concentrations (Sad., Se., and Sd == Sdegr)'
13. Monod kinetics in CSTR (cf. Sect. 6.1.1). The following set of balance
equations were used (see Equ. 6.3):
rx = jl(s)x - D .x (II.39)
1
- rs = y,- jl(s)x + D(so - s) (II.40)
xis
with Monod-type kinetics according to Equ. 11.4. Whereas Fig. 6.1 showed
the typical plot of "chemostat" behavior (steady-state concentrations x
and "8 versus dilution rate D), Fig. 11.33 represents a similar plot with
variations in initial substrate concentration So. The "washout state" is
indicated (1 x, 1"8) together with a second stationary state characterized by
a nonvanishing biomass concentration 2 X, 2"8.
Whereas for batch cultures the final concentrations depend on the initial
concentrations, in an open reactor like the CSTR "equifinality" is estab-
lished, where the end value of stable concentration x is independent of
initial values, as shown in Fig. 11.34 in case of jlmax = 0.8 h-1, Ks = 10
mg '1-1, YxlS = 0.5, So = g .1-1 (from ZIMET, Jena).
CSTR behavior in case of more complex kinetics has been shown in

Figs. 6.4 through 6.11 and 6.14 through 6.20. Alternative bioreactor opera-
tions (CPFR, NCSTR, RR, etc.) were represented in Sects. 6.4 through 6.8.
BIBLIOGRAPHY
General:
Ropke, H., Riemann, J. (1969). Analogcomputer in Chemie und Biologie, Berlin:
Springer-Verlag.
Knorre, W.A. (1971). Analogcomputer in Biologie und Medizin. Jena: VEB G. Fischer.
Romanovsky, J.M., Stepanova, N.V., and Chernavsky, D.D. (1974). Kinetische
Modelle in der Biophysik. Jena: VEB G. Fischer.
Levin, S. (ed.) (1981). Modeles Mathematiques en Biologie, Berlin: Springer-Verlag.
Knorre, W.A. (1980). Kinetische Modelle in der Mikrobiologie in "Biophysikalische
Grundlagen der Medizin (Beier W., Rosen R., eds.) Stuttgart: Fischer.
Spain, J.D. (1984). Basic Microcomputer Models in Biology, Addison-Wesley Pub!.
Comp., Reading, Massachusetts, USA.
Rogers, P.L. (1976). In Advances Biochem. Engng. 4,125.
Special:
Bajpaj, R.K., Reuss M. (1981). Biotechnol. Bioengng., 23, 717.
Moser, A., Schneider, H. (1988). Bioprocess Engineering 3, in press.
Theophilou, J., Wolgbauer, 0., and Moser, F. (1979). Gas-Wasser-Fach-Wasser/
Abwasser 120, 119.
APPENDIX III
Microkinetics: Derivation of Kinetic
Rate Equations from Mechanisms
The objective of this appendix is to demonstrate the concepts of the rds

(rate-determining step) and qss (quasi-steady-state). These are both of great
importance in kinetic modeling, as explained in Sects. 204, 4.2, and 5.2. At the
same time, some well-known approaches to the microkinetics of enzyme
reactions are presented here; these are more fully discussed in sect. 5.2.2.1.
IlL 1 Simple (Chemical) Reaction

Reaction scheme:
(1II.1 )
Reaction mechanism (collision theory):
2A~{A} +A (III.2a)
A + {A}~2A (HI.2b)
{A}~B (HI.2c)
Rates of individual steps:
r1 = kl d. (II I.3 a)
r2 = k 2 c A c{A} (lII.3b)
r3 = k3 c{A} (II I.3 c)

The qss concept (assuming that r{A} = 0) gives
riA} = kl . c1 - k2 . CA . CIA} - k3 . CIA} = 0 (lIlA)
This assumption enables the derivation of an equation for the activated state
{A}, which normally cannot be measured:
(111.5)
111.3 Briggs-Haldane Approach 437
Now, the rds concept is applied, assuming that

rtot = r3 = k3 . C{A} (III.6)
resulting in an expression for the global rate rtot by using Equ. III.5
kl ci
rt t = (III. 7)
o (k2/k3)CA + 1
In case of CA 0, the final form of a kinetic equation is
(III.S)
which is of simple first order.
III. 2 Michaelis-Menten Approach (cf. Equ. 2.53)

Reaction mechanism:
S+E~{ES}~E+P
k_1
(III.9)
rds concept:
(III.10)
qss concept applied for {ES}:
dC(ES}
dt = k+1's'e - Ldes} = 0 (II1.ll)
result in Equ. 2.53 with
(III.12)
and
(III.13)
Remember that e = eo - {es}.
III. 3 Briggs-Haldane Approach

The same approach as in Equs. III.9 and III.10 but with a qss concept
including a steady-state assumption of all k i values gives:
(II1.l4)
The result is the same type of a kinetic expression (cf. Equ. 2.53) but with a
438 III. Microkinetics: Derivation of Kinetic Rate Equations from Mechanisms
different meaning of the saturation constant Km:
K = -Ll + k+2
--c--- (IILI5)
m k+1
IlIA Langmuir-Hinshelwood Approach

Reaction mechanism:
E + s ~{ES} ~ {EP} ':"=:-E + P

k~ k~
(IILl6)
Rates of individual components with qss:

ds
dt = -k+l se +L 1 {es} (III.17a)
de
dt = -k+l se + Ll {es} - L3 e p + k3{ep} (I1I.l7b)
d{ es}
--;[t = k+l . s e - Lr{es} - k 2{es} = 0 (111.17c)
d{ep}
----;[t" = k+2{es} + L3{ep} - k+3{ep} (III.17d)
dp
dt =k+3{ep} - L 3 ep (III.17e)
with
e + {es} + {ep} = eo (IILl8)
and steps 1 and 3 being at equilibrium (Ks = adsorption, Kp = desorption
constant):
(III.19a)
{ep} L3
--=-p=Kpp (III.l9b)
e k+3
Equation 111.18, rewritten by using Equ. IILl9a, b gives

{es}(l + Kss + Kp p)
= eo (III. 20)
Kss
which can be replaced in Equ. IILl 0 (rds!) resulting in the final form of a kinetic
term:
Kss
r =r (Hl.2l)
tot max 1 + K s . s + K p . P
III.6 Reversible Langmuir-Hinshelwood Approach 439
This is the general form of the Langrnuir-Hinshelwood equation applicable

to heterogeneous chemical catalysis.
IlL5 Reversible Michaelis-Menten Type

Reaction mechanism:
E+S~{ES}~E+P
k-, k-2
(111.22)
The rds concept yields

dp
r tot = dt = k+2{es} - k_ 2 ep (111.23)
The qss concept gives

. S + L2p)e O
{ es } = - - - -(k+l
-''----'''---=---- (111.24)
k+l . S + Ll + L 2P + k+2
Thus
(111.25)
and in equilibrium, where

k+l . k+2 . S = Ll . L2 . P
and using the definition of Keq (cf. Equ. 5.22) resulted in Equ. 5.21.
Simplifications are achieved if P = 0 and s = 0 for the initial rates of forward
and backward reaction, which are used in enzyme kinetic studies.
IlL6 Reversible Langmuir-Hinshelwood Approach

Reaction mechanism:
E + s ~{ES}
k_,
2.h{EP}
~
~E
~
+P (III.26)
The rds concept states that

rtot = k+2{es} - L2{ep} (III.27)
and qss concepts are needed for both complexes {ES} and {EP}:
r{es} = k+l -s-e - Ll res} + L2{ep} - L2{es} = 0 (III.28a)
r{es} = k+2{es} - L2{ep} + k-3 e p - k+3{ep} = 0 (III.28b)
In analogy to the preceding case (cf. Equs. 111.18 and II1.19) expressions for
both complexes are written as
440 III. Microkinetics: Derivation of Kinetic Rate Equations from Mechanisms
eo s Ks
{es} - ----'----"-- (III.29a)
1+Ks 's+K p 'p
eO'p'K p
{ep} - ---=--=---=-- (III.29b)
1+Ks 's+K p 'p
and Equ. III.27 can be rewritten by using Equ. III.29a and b as
k+2'S'Ks L 2'P'K p )
rtot = eo ( - (III.30)
1+Ks 's+K p 'p 1-Ks 's+K p 'p
resulting in an expression already shown in Equ. 5.23. Very often, the resulting
rate equation has the sameJorm as Equ. 5.21 or Equ. 5.23 (both of them can
be brought in a similar form) but with a different interpretation of the kinetic
parameters. There are many mechanisms more complicated than those de-
scribed here that nonetheless generate the same type of formal kinetics, for
example, Equs. 2.54, 5.21, or 5.23.
Index
A B
absolute deviation, 89 Back mix-plug flow model, 84
absolute rate, 19,20 balance, mass, 118ff
absorption, 221 balanced growth, 145, 157, 272
activated state, 203, 436ff balanced system, 11, 116, 145, 272
activation, 199 balancing method, 53, 382ff
activation energy, 199, 200 batch reactor see discontinuous reactor
activity function of product formation, bed expansion, 117
276 bed porosity, 369
adaptation constant, 277 bench scale (lab scale), 44
adaptational parameter, 60 Bingham plastic, 385
adaptive modeling, 50 biocoenosis, 10, 259, 260, 293
adsorption, 240 biodisc reactor, 68, 139
aeration number, 106 biofilm,68
age, 246 biofilm kinetics, 151, 283ff
allosteric control, 217 biofilm operation, 69, 139
allosteric inhibition, 232 biofilm reactor operation, 69, 139, 358ff
allosterie, 212, 217 biological inertia, 198, 212, 214
amensalism, 261, 265 biological rate equation, 178, 283f
anhydrase-method, 190 biological test system, 41, 87, 90, IlOff
apical growth, 238 biomass, 19
apparent morphology, 390 biomass holdup, 367
apparent rate, 172 bioparticle terminal Re number, 369
apparent value, 172 Bioprocess design, 307ff
apparent viscosity, 387 Bioprocess kinetics, 19, 197ff
Arrhenius equation, 199 Bioprocess technology, 5, 7, 13
Arrhenius plot, 200, 201 bioreactor, 2, 66ff, 73, 307
assimilation, 220, 221 bioreactor concept, 112
Atkinson equation, 283 bioreactor dimensioning, 127
ATP, 29, 30, 245, 282 bioreactor model, 44, 45, 56, 112, 118ff,
attraction domains, 320 125,307
autocatalytic process, 344, 345 bioreactor operation, 112, 307
autocatalysis, 25 bioreactor performance, 307ff
autoregulation, 205 biosociety, 1
available electrones, 31 biosorption, 238, 342
axial dispersion, 122, 175 biosorption model, 238, 433
442 Index
biosorption rate, 434 closed environment, 261

biotechnology, 1 C-mole (of biomass), 28
Biot-number, 187 COD, 292
bistable system, 319 community matrix, 261
bistability, 319, 320 commensalism, 261, 265
blads box model, 48 compartment 279, 144, 145
Blackman-kinetic, 218 compensation, 203
bleed stream, 378 competition, 261, 262f
BOD, 292 competitive inhibition, 209
BODs, 292 completely mixed microbial film fermen-
BOD pu 292 ter, 128
Bodenstein-number, 75, 80, 86, complex reaction system, 406
347 computer simulation, 412ff
Boltzmann-constant, 203 COiNaOH-method, 190
bottleneck, 206, 436ff, 218 concentration polarization, 373
boundaries (closed, open), 76 concept of characteristic times, 14lff
boundary conditions, 76, 169, 176 conductivity, 81, 93
branching rate, 238 conservation of mass, 22
Briggs-Haldane approach, 437 consistency index, 387
Briggs-Haldane equation, 208, 437 constant of adaptation, 277
bubble column, 67 constant growth phase, 219, 244
bulk, 18, 168 constant-volume continuous stirred tank
reactor, 119
C constant volume reactor operation, 113,
calculation methods, 349ff 119
calorimetry, 104 consumption activity coefficient, 243
Carbon-flow-brenching concept, 37 continuous mode of reactor operation, 10,
Carman-factor, 147 113
cascade, 77, 113, 333, 353 continuous recycle reactor, 123, 351
cascade of reactors with cell recycling, continuous plug flow reactor (CPFR),
353 113, 121, 337ff
Casson equation, 386, 389 continuous stirred tank reactor (CSTR),
Casson viscosity, 386 113, 119, 308ff, 434
catabolite repression, 145, 251, 237 Contois-kinetics, 217
catalytic constant, 234 controlled filmthickness, 126, 138
catalytic process, 344 convection coefficient, 91
cell age, 246 convection theory, 91
cell age (mean), 246 conversion, 38
cell cycle, 379 costs, 40
cell model, 58, 77 critical diameter, 147, 150, 177
cell tissue culture, 67 critical film thickness, 147, 150, 177
characteristic diameter, 105 critical S-concentration, 252
characteristic length, 75 critical treshold concentration, 233
characteristic times, 142, 393 critical washout point, 309
characteristic-time concept, 14lff cross-inhibition, 255
chemostat, 119 crowding factor, 391
circulation time distribution, 88 CTR control, 309
circulation time, 82, 86, 89, 90 cube root equation, 289
closed boundaries, 76 cycle time distribution, 87, 88
closed environment, 261 cyclic operation, 116, 325
Index 443
D DIO value, 202, 294

Damkoehler-number-l" degree, 347 dynamic flow eqUilibrium, 204
Damkoehler-number-2nd degree, 148, dynamic method, 90, 92, 94, 102
176, 185 dynamic models, 95, 272ff
damped oscillations, 270 dynamic process (kinetics), 274ff
Danckwerts boundry conditions, 169 dynamics of enzyme synthesis, 213
Danckwerts reactor, 129, 191
death, microbial, 227
death, phase, 225 E

death rate constant, 227, 228, 230 Eadie-Hofstee plot, 166, 174, 180
decay rate (decomposition), 244 economy (of 2), 10 I
decimal reduction time, 202 economics, 38
declining growth phase, 243 effective hyphal length, 238
deduction, 10 effectiveness factor, 11, 148, 170, 182,
deductive method, 50 186, 286
definition of reaction rate, 23ff, 155 effectiveness, GIL, 148
degradation (substrate), 221, 239,240 effectiveness, intraparticle, 148
degree of Hinterland, 99 effectiveness, liquid/particle, 149
degree of (micro) mixing, 81 effectiveness, surface based, 182
degree of 02-utilization, 99 effectiveness, volume based, 170, 178
degree of reductance, 32 effective rate, 170, 173, 177, 179, 180,
degree of reduction, 32 185
degree of segregation, 72, 83, 349, 350 effectivity, 170
depth of penetration, 179 efficiency general, 11, 47
deviation variables, 275 efficiency of motor, 100
dialysis (membrane) reactor operation,
efficiency of 2 , 101
371 electrode dynamics, 97
dialysis reactor, 371 electrode response, 97
diauxic growth, 237, 251, 431 elementary analysis, 25, 382, 406
diauxie, 226 element balance, 382
differential analysis, 11, 156 element-species matrix, 407
differential evaluation method, 154 elimination, 221, 239, 240
differential reactor, 11, 152 elutriation, 117
diffuson, 22, 75, 91 empiric pragmatic approach, 11, 16
diffusion regime, control, 171, 177, 185, endogeneous metabolism, 198,216,
189 228ff
dilution rate, 119 energy dissipation, 147, 181
"direct linear" plot, 167 energy efficiency coefficient, 31
discontinous mode of reactor operation, energy of activation, 199
113,119 engineering morphology, 390
discontinuous recycle reactor, 123 enhancement factor, 170, 189
discontinuous stirred tank reactor, 119 enhancement of transport, 170, 188ff
dispersion (coeff.), 75, 122 enhancing substrate, 255
dispersion model, 74, 122 enthalpy, 203
dissipation (energy), 147, 181 enthalpylentropy compensation, 203
distributed parameter reactor, 121, 113, entropy, 203
151 enzyme kinetics, 436ff
Dixon-plot, 211 enzyme mechanisms, 436ff
double-substrate-Iimitation function, 255ff enzyme reactor, 68, 311, 338, 341, 344
D-value, 294 enzyme regulation and control, 211
444 Index
equifinality, 205, 321, 434 G

equilibrium constant, 57, 437ff gas analysis, 90, 98
equivalent stage, number, 77 gas/liquid reactor model, 357
essential substrate, 255 gas phase dynamics, 96
exluded-volume-concept, 389 gas hold-up, 91
expansion index, 368 gas-in method, 91
exploratory research, 44 gas-out method, 91
exponential growth phase, 225, 243 Gates (MarBar) plot, 163, 164
extended fed-batch culture, 325 gau f3-elimination method, 407
extent of reaction, 406, 411 G-compartment, 145, 279
external transport limitation, 170, 171ff, generalization, 52, 222, 233, 236, 254,
183ff 363
extractive fermentation, 377 glucose oxidase method, 91
glycolysis oscillations, 205, 206
Gompertz equation, 289, 290
F gradientfree reactor operation, 153
falling film reactor, 128 gradientless reactor, 153, 157
falling jet reactor, 128 gradient-reactor, 157
fast reaction, 189, 190 graphical methods, 156
fed-batch, 113 gross stoichiometric equation, 26, 382,406
fed-batch cycle, 113 growth, 21, 23, 25, 216
fed-batch reactor operation, 114ff, 325ff growth association, 241, 242, 244
feed rate, 114 growth-enhancing substrates, 255
fermentation enthalpy, 249 growth kinetics, 216ff
fermenter see bioreactor growth of pellets, 288
Fick's law of diffusion, 22
filamentous growth, 237
film thickness, 91 H
filtration time, 391 Haldane-relationship, 208
first order kinetics, 214, 216, 239, 253 half-order reaction, 179, 185,214,287,
fitting parameter, 60 369
fixed bed reactor model, 360 Hatta-number, 189
flow behavior index, 385 heat balance, 103
flow rate, 114 heat capacity, 103
fluidization model, 368 heat, general, 101
fluidization velocity, 117 heat formation, 103
fluidized bed, 117, 128,366 heat of combustion, 250
fluidized bed biofilm reactor (FBBR), heat of fermentation, 103, 249, 250
117, 366[f heat production, 103, 247ff
food technology, 294 heat transfer coefficient, 20, 104, 105
formal kinetic evaluation method, 60, heat transfer rate, 20, 101ff
242 Henry-distribution coefficient, 92
formal kinetics, 59 Henry-equation, 160
formal macro-approach, 46, 59 heterogeneous bioprocesses, 151, 168
Froude number, 108 heterogeneous kinetics, 151, 283 ff
frozen system, II, 116, 145,272 heterogeneous reactor, 9, 70, 139
Fujimoto-equation, 217 heterogeneous system, 70, 139, 168
fungal growth, 237 HilI-coefficient, 212
F-value, 294 Hill-equation, 212, 217
Index 445
HiII-kinetics, 212 K
HiII-plot, 213 K-compartment, 145
Hinshelwood's network theory, 238 key reaction, 279, 409
hinterland, 99, 189, 190, 358 key variable, 26, 408
homeostasis, 261 kinetic model, 54ff
homogeneous kinetics, 151,216 kinetic modeling of lag-phases, 225, 430
homogeneous rate equations, 151, 216ff kinetic modeling of endogenous metabo-
homogeneous reactor, 9, 70 lism, 227, 431
homogeneous system, 70 kinetic regime, control, 171, 177, 189,
horizontal position of reactor, 126 190
hyperbolic type, 212 kinetic similarity, 45
hyphal growth unit, 239 KLa-value, 94, 106
Kolmogoroff-theory, 147, 181
Konak-equation,222
I Kono approach, 219
ideal single-stage reactor, see CSTR Kono concept, 219
idiophase, 378 Kono-kinetics, 219
inactivation of spores, 294 Kozeny-Carmen equation, 391
incomplete mixing, 313
incomplete substrate penetration, 179
induction, 145, 213 L
induction phase, 243 lag-phase (growth), 145, 216, 225
inductive method, 50 lag-time, 145, 214, 216, 225, 275
ingestion, 220 Langmuir-Hinshelwood approach (revers-
inhibition, 145, 198, 209, 233, 252, 255, ible), 208, 439
339 Langmuir-Hinshelwood kinetic (irreversi-
inhibition types, 209, 210, 233, 234 ble), 208, 438
inhibitor, 209 Langmuir-kinetics, 57, 160, 165, 166,
inhomogeneity, 72, 82, 83 240
initial rate, 208 Langmuir plot, 166, 181
integral analysis, 11, 154, 156 level of sterility, 293
integral evaluation method, 155 limitation-substrate, 16, 18
integral reactor, 11, 151, 338 limitation-transport, 170
integrated bioreactor system, 68 limit cycle (oscillation), 269
integrating strategy, 41ff, 53, 38lff, 396 limiting viscosity number, 392
interactions, 43,61, 140,261,385 Lineweaver-Burk plot, 166, 180, 209
interactive model, 256, 257 linear dependence, 26, 409
internal transport limitation, 170, 175ff, linear growth, 290, 376
183ff linearization, diagram, 159, 16lff
intrinsic rejection, 373 linearly independent stoichiometric equa-
intrinsic viscosity, 392 tion, 409
ionic strength, 93, 232 lineary independent equations, 409
invariance, 215 local isotropic turbulence, 147
isokinetic temperature, 203 logarithmic mean value, 98, 103
logistic equation (low), 226
log-mean value, 98, 103
J longitudinal dispersion, 122
jeopardy, to put a model in, 52 Lotka-Volterra analysis, 268
joint analysis, 93, 167 Lotka-Volterra model, 269
446 Index
Lotka-Volterra relationship, 220, 269 Michaelis-Menten type (irreversible), 208

lumped parameter reactor, 119, 151 microbalances, 25
lysis, 312 microbial death phase, 225, 227
microbial interactions, 259
microbial product formation, 240
M microfiltration, 371
macrobalances, 25, 54 microkinetics, 45, 55, 204, 436
macrokinetics, 45, 56, 139 micromixing, 72, 81, 350
macromixing, 74, 350 minimal medium, 16
macroscopic principle, 18,46, 382 mini lab reactor, 16,44
macroscopic yield, 28 mixedness, 71
Maillard reaction, 295 mixed plug flow, 87
maintenance coefficient, 229, 230, 282, mixed population, 259
312 mixed population kinetics, 259ff
Malthus type, 224 mixed substrate kinetics, 250ff
mass-action law, 143, 2r1 mixing, 70, 81
mass balance, 118ff mixing decay rate constant, 88
mass flux, 22, 373 mixing models, 84
mass loading rate per unit biomass mixing number, 105, 109
(sludge), 275 mixing time, 81, 83, 89
mass transfer, 20, 169 model, 49
mass transport with simultaneous reac- model building, 50, 51, 52
tion, 169 model discrimination, 52
master reaction, II, 206, 218, 436ff model identification, 50, 52
mathematical model, 49, 168 model parameter, 50
mathematical modeling, 48, 49 model reactor, 84, 128, 129, 191, 130,
matrix, 407 395
matrix of the stoichiometri coefficients, modulus, 149, 172, 176, 182, 186, 188
407 mole of microorganisms, 27
maturation time, 246, 346 momentum method, 95, 96
maximum mixedness, 71 Monod equation (kinetics), 56, 160, 166,
mean diameter, 91, 181, 182 216,217,218,224,412
mechanism, 55 Monod kinetics with death rate, 227,
mechanistic kinetics, 55 228,431
mechanistic model, 55, 58 Monod kinetics with lag time, 225, 226,
medium design (optimization), 16 412
membrane performance, 372 Monod kinetics with S-inhibition, 216,
membrane permeability, 372 232,431
membrane reactor, 371 Monod with O2 limitation, 256, 432
metabolic process enthalpies, 249 Monod type kinetics with endogeneous
metabolite repression, 145, 252, 237, 389 metabolism, 228, 431
methodology, 13, 385 monostability, 309
Michaelis-Menten approach, 437 morphology, 389
Michaelis-Menten equation, 56, 208, 437 morphology factor, 389, 390, 391, 392
Michaelis-Menten kinetics, 56, 160, 437 morphology index, 389
Michaelis-Menten kinetics (reversible), Moser equation (kinetics), 217
208,439 multicomponent system, 294
Michaelis-Menten mechanism, 437 multi inhibition model, 432
Michaelis-Menten model, 437 multi-loop mixing model, 85
Index 447
multiple phase bioreactor model, 124 oxygen economy, 10 1

multiple steady states, 265, 267, 321 oxygen efficiency, 101
multi-phase reactor model, 124 oxygen saturation, 92
multi-purpose reactor, 128 oxygen solubility, 92
multiresponse analysis, 167 oxygen transfer number, 106, 109, 110
multistage reactor system, 329ff oxygen transfer rate, 90, 125, 144
multistage system, 329, 330 oxygen utilization degree, 99
multistage system with recycle, 329 oxystat, 309
multistream multistage operation, 329,
334
multistream system, 329 p
multisubstrate kinetics, 250ff parameter, 50
mutualism, 261, 266f parameter estimation, 50, 15lff
mycelia, 237, 389 parameter sensitivity analysis, 51, 414ff
mycelial (filamentous) growth, 237, 389 parasitism, 261
partial dilution rates, 335
Pected number, 75
N pellet, 193, 288, 359
neutralism, 261 penetration, 221
Newtonian fluids, 385 penetration depth, 129, 179
nomenclature, 5, 19,21 penetration theory, 91
noncompetitive inhibition, 209 perfect bioreactor, 41, 44, 126ff, 191
non-growth-Iinked product formation, performance, 307
242, 244, 246 performance equations, 307ff
noninteractive model, 258 peripheral zone model, 289
Non-Newtonian fluids, 385 periodic eqUilibrium curve, 333
non-stationary kinetics, 272ff periodic reactor operation, 116
non-stationary reactor operation, 411 period of oscillation
nonvanishing biomass, 320, 434 permeability, 372, 374
non-wash-out state, 320 permeate, 373
normal catalytic process, 344 phased culture techniques, 379
number of equivalent stage, 77 photo bioreactor, 238
number of tanks in a cascade, 77 photosynthesis, 237
numerical fitting, 57, 201, 237 pH-auxostat, 309
pH-dependence (kinetics), 237
pH-stat, 309
o phyto technology, 1
observed rejection, 373 pilot plant (reactor), 16, 44
"on-off' concept, 252 pilot scale, 16, 44
open boundaries, 76 Planck-constant, 203
open environment, 261 plug flow reactor, 113, 121, 337f
optimum, biological, 198 plug flow reactors with recycling, 354
optimum conversion, 39, 339, 341, 350 point, 72
ordinary sinetics, 145 polarization, 373
oscillation, 206, 266, 269, 270, 273 point, 72
OTR control, 309 Powell kinetics, 217, 287
OUR control, 309 power consumption, 100
overall transfer coefficient, 186 power index, 385
overlapping (substrate), 252 power law constant, 385
448 Index
power law fluid, 385 R

power number, 105 rank of matrix, 407
predation, 261 rate, 20
predator, 268 rate coefficient, 20
predator-prey interactions, 268ff rate determining step (rds), 11,42,218
preexponential factor, 199 rate of apical growth, 238
pressure, 10 1 rate of branching, 238
prey, 268 rate of permeation, 373
principle of separation, 47 "rds concept", 11, 206, 436ff, 272, 360
process analysis, 139ff reaction enthalpy of fermentation, 249
process design, 7, 44 reaction order, 214
process kinetics, 52, 138 reaction rate definition, 23, 24, 25
process kinetic analysis, 11, 42, 44, reaction time, 92, 146
138ff reactor with gradients, 157
process parameter, 51 real plug flow reactor, 122
process variable, 19, 52, 60 real stat , 119
product formation (kinetics), 240ff, 276f, recycle rate, 79
314f recycle ratio, 79, 86
product formation activity function, 276 recycle reactor operation, 79, 154, 351ff
product inhibition, 145, 316f reductance, 32
product inhibition kinetics, 216, 234ff reductance degree, 32
production of secondary metabolities, reduction, 32
241,244,246,247,278,388,432 reference fermentation, 112
productivity, 38, 309, 353 regime analysis, 14lff
profit, 38 regulation of enzyme amount, 145, 206ff,
proto cooperation, 266 211
pseudo-differential reactor, 153 regulation of enzyme activity, 145, 206ff,
pseudohomogeneity, 11, 146 211
pseudohomogeneous L-phase reactor Reith-approximation, 189
model, 169 relative growth rate, 23, 222
pseudohomogeneous rates, 11, 237, 240, relative rate, 20
285,360 relati ve velocity, theory, 147
pseudo-integral reactor, 152 relaxation time (see characteristic time),
pseudokinetics, 139, 290f 277
pseudokinetic parameters, 173 relaxed steady state, 116
pseudokinetic phenomena, 173 repeated fed-batch, 114, 325
pumping capacity, 393f, 17 repression, 145, 213
repression kinetics, 237, 248, 389
residence time distribution, 74, 341
Q response time, 95, 142
Q-value, 276 reversed-two-environment model, 84
qe,24 reverse osmosis, 371
qH,24 Reynolds-number, 108
qQ, 24, 94, 139, 223 rheology, 385
qs,24 ribosome, 214, 252
quantification, 6, 18,47,73,241
qss (quasi steady state), 11, 42, 120,
157 S
quasi-steady state reactor operation, 120, salinity, 223
327 saturation constant, 56
Index 449
saturation-type kinetics, 56, 160ff, 198, stable steady states, 309, 319, 323
217ff standard bioreactor, 127
Sauter-diameter, 91, 182 stationary method, 98
scale-down, 45 stationary phase of growth, 225, 226
scale-up, 17, 381 stationary state, 119, 157, 204, 207
scale-up concept, 381 steady state, 119, 157,204,207,320,
scale-up correlation, 107ff 380
Schmidt-number, 108 steady state growth, 157, 273, 274
screening, 14 sterilization process, 346
second order kinetics, 214 sterilization kinetics, 292ff
secondary metabolite productions, 241, sterilization level, 293
345,432 sterilization process, 293
segregated kinetic model, 49 sterilizer, 68
segragation, total, 72 stimulation, 198
selectivity, 38 stoichiometrical coefficients, 34
semibatch culture, 113, 325, 4-10 stoichiometrical dependence, 26
semicontinuous mode of reactor opera- stoichiometric line, 256
tion, 10, 113, 410 stoichiometric matrix, 407
semicontinuous stirred tank reactor, 113, stoichiometry, 11,20,25,34, 382ff,
119 406ff
sensitivity analysis, 51, 53, 414ff stoichiometry (balancing methods), 382
separation, 47 stoichiometry of complex reactor opera-
sequential S-utilization, 25lff tion,406ff
shear rate, Ill, 385, 387 strain gauge, 100
shear stress, 385 strategy, 15, 396
Sherwood-number, 108, 174 structured cell model, 58, 145
shift-down, 319, 322, 323 structured kinetics models, 145, 278ff
shift-up, 319, 322, 323 structured modeling of bioreactors
sigmoidal type, 212 (OTR), 144, 392
significant variable, 18 structured models, 82, 87, 144
simplification, 46 structured reactor models, 82, 87, 144,
simultaneous S-utilization, 253ff 392ff
single cell model, 58 substrate, 145
single-stream multistage operation, 331 substrate inhibition, 145
single-stream reactor operation, 331 substrate inhibition kinetics, 216, 232ff,
sloping plane bioreactor, 126, 128 318
slow reaction, 171, 189 substrate-utilization kinetics (sequential),
sludge adsorption, 238 251
sludge loading, 275 substrate-utilization kinetics (simultane-
sorption capacity, 240 ous), 216, 253
sorption number, 106, 109, 110 sugar-feeding strategies, 433
specific growth rate, 23 sulphite oxidation, 189ff
specific heat capacity sulphite oxidation method, 90, 189ff
specific rate, 20 surface area, 90
spores, 293, 295 surface renewal rate, 91
spouted bed, 117 surface renewal theory, 91
spouted fluidized bed, 117 survival of the fittest model, 262
spread of distribution, 76 symbiosis, 261
stability analysis, 266, 318 synergism, 261
stable oscillations, 269, 270 synchronous culture operation, 378ff
450 Index
synchronous/synchronized pulsed and troprophase, 378

phased culture, 379 tubular reactor, 67, 69, 238
systematic approach, 5, 15, 41 turbidistat, 309
two-compartment model, 144, 280ff, 394
two-environment model, 84
T two-film theory, 91
tanks in series model, 77 two-phase reactor model, 357
Teissier kinetics, 217, 232 two-region-mixing model, 84
temperature, 198 two-stage reactor, 330
temperature dependence, 198 two-zone model, 84, 192
test of pseudohomogeneity, 146ff types of kinetic models, 54, 151
test system of biolog. test system
theory of activated complexes, 203
thermodynamic efficiency, 31, 32, 33 U
Thermodynamics, 25 Uhlich approximation, 103
Thiele-modulus, 149, 172, 176, 178, ultrafiltration, 371
182, 187, 286 unbalanced growth, 157
Thiele-modulus based on biofilm thick- uncompetitive inhibition, 209
ness, 188 uncontrolled film thickness, 138
Thiele-modulus for zero-order, 179, 186 unified performance, 364
Thiele-modulus generalized, 187 unsegregated model, 49
Thiele modulus f. 1st order, 178, 187 unstable state, 319
thin layer (tubular film) fermenter, 126, unsteady state reactor operation, 157
128, 129 unstructured models, 118, 146, 216
three-compartment model, 145 unstructured models (kinetics), 146,
three-constant (parameter) equation, 218, 216ff, 274ff
235 unstructured models (reactors), J 18ff
threshold concentration, 233
time lags, 97
torque, 100 V
Toe, 292 variable, 52, 60
total hyphal length, 390 variable volume reactor operation, 119,
total segregation, 7lf 325
toxic power number, 236 variance, 72, 89
toxicity, 198 Verhulst-equation, 224
trajectory, 264, 321, 320, 323 Verhulst-Pearl's equation, 224, 225
transient behavior of the CSTR, 157, 321 viscosity, 385
transient operation, 157 volume-based effectiveness factor, 178,
transient reactor operation, 157, 325 182
transient kinetics, 145, 321 volumetric heat, 18, 20, 103
transient state, 207, 321 volume-utilization (see hinterland), 99
transient phase, 243
transition line, 257
transport coefficient, overall, 185ff W
transport coefficient, volumetric, 94 Walker-diagram (plot), 161,253
transport enhancement, 140, 170, 188ff wall growth, 313
transport limitation, 140, 170 wash-out, 309
transport rate, 19 wash-out point, 309
Index 451
wash-out state, 309, 320 yield constant, 20, 28, 29, 35ff, 86, Ill,
waste water plant, 2, 66 223, 230, 242, 312
waste water treatment, 10,67, 260 yield factor, 20, 28, 29, 35ff, 86, Ill,
water activity, 198 223, 230, 242, 312
whirling bed, 117 yield stress, 20, 28, 29, 35ff, 86, Ill,
Williams model, 145,280 223,230, 242, 312
working principles, 46
Z
y Z-value, 202, 294
yield, 20, 39 zero order kinetics, 214, 216, 239
yield coefficient, 20, 28, 29, 35ff, 86, Zootechnology, I
Ill, 223, 230, 242, 312

Bioprocess Technology Kinetics and Reactors by Professor Dr. Anton Moser

Hochgeladen von

Dokumentinformationen

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

Bioprocess Technology Kinetics and Reactors by Professor Dr. Anton Moser

Hochgeladen von

Copyright:

Verfügbare Formate

Bioprocess Technology

design &. development

Engineering Disciplines in Biotechnologies

With 279 Illustrations

Library of Congress Cataloging-in-Publication Data

Printed on acid-free paper.

Revised and expanded translation from the German edition

Typeset by Asco Trade Typesetting Ltd., Hong Kong.

ISBN -13:978-1-4613-8750-3 e- ISBN -13 :978-1-4613-8748-0

This book is based on a 1981 German language edition published by Springer-

This book attempts to present the formal kinetic macro-approach following

Graz, Austria ANTON MOSER

Chapter 1 Introduction ................................... .

Chapter 2 The Principles of Bioprocess Technology ........... . 13

Chapter 4 Process Kinetic Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . 138

4.4.3 Results of Differential and Integral Analysis: Linearization

Chapter 5 Bioprocess Kinetics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197

5.7.1 Definitions of Balanced Growth and Steady-State Growth ........ 272

Chapter 6 Bioreactor Performance: Process Design Methods ... . .. 307

6.9.3 Influence of Mycelium-The Morphology Factors {"Apparent

Appendix I Fundamentals of Stoichiometry of Complex

Index..... .... .............. .... .. . . . ...... ...... ........ .. 441

BOD [kgm- 3 C I ] biological oxygen

C [kg] amount of CO2

CTD circulation time

D [m 2 's- l ] diffusion or dispersion

E [m 2 's- 1 ] convection coefficient

[kg O 2 . kWh- 1 ] oxygen economy

[m 3 h- 1 ] volumetric flow rate

AG [J. mole- 1 ] free reaction enthalpy

K [kgm- 3 ] constant of equilibrium

kE [S-l J rate constant of

L em] length (general term)

nB bed expansion index

P [kg] amount of product

S [kg] amount of substrate

2S [kgm- 3 J "non washout state" S

y, 1';u yield coefficient (general

Z value (Equ. 5.9)

ads, a adsorption B bulk or bubble or

calc calculated L liquid phase

s sensitive COD chemical oxygen

z coordinate ES enzyme substrate

FBBR fluidized bed biofilm

K K-compartment in rCSTR continuous stirred tank

Ys,Yx degree of reductance (reduction)

1.1 Biotechnology: A Definition and Overview

FOOD BEVERAGE WASTE WATER-,WASTE- TISSUES-

H I 0 LOG I CAL REA C TOR S

cheese, for example, would be considered as bioprocesses of food technology.

-... - ...... --.:..-

FIGURE 1.2_ Predicted development of biotechnical processes (---) compared with

1.2 Bioprocess Technology

and more the problems involved in integrated quantification of bioprocesses.

FIGURE 1.3. Areas of biotechnology potentially contributing to the development of

process control including techniques for carrying out the process on an

TABLE 1.1. Comparison ofbioprocessing and chemical processing.

TABLE 1.2. Process engineering comparison between simple fermentation and

From A. Moser, 1981.

uous, semidiscontinuous, or fully continuous operations; unstable behavior

2.1 Empirical Pragmatic Process Development

2.1.1 PRODUCTION STRAINS

2.1.2 THE STARTING POINTS

In general, four situations are conceivable in starting a program of process